refactor(chat): route browser chat through one runtime

Co-Authored-By: Claude Opus 4.8 <[email protected]>
Claude-Session: https://claude.ai/code/session_01ESFAnh2t9HmLwng8oW95St

.woodpecker/ci.yml

@@ -109,6 +109,16 @@ steps:
       # `apk add` guarantees openssl is present on PR pipelines too (and is a
       # fast no-op once the rebuilt image already ships it).
       - apk add --no-cache openssl
+      # Pi runtime (Invariant R): invariant_r_unittest.py hard-requires an
+      # installed `pi` binary at exactly this measured version — the test
+      # boots Pi's real tool registry to prove the read-only carve-out
+      # resolves to real, unshadowed builtins, and fails loud (by design)
+      # when the runtime is absent or drifts. The canonical Pi is
+      # @earendil-works/[email protected] exactly (@mariozechner/* is
+      # embedded-legacy). Step-level install because ci-base image publishes
+      # are currently blocked on registry auth; fold into Dockerfile.ci once
+      # that is fixed, keeping this as a fast no-op guard.
+      - npm install -g @earendil-works/[email protected]
       # postgresql-client (pg_isready) is baked into ci-base.
       # Wait up to 60s for CI postgres to be ready; fail fast if it never comes up.
       - |

AGENTS.md

@@ -11,48 +11,87 @@
 
 ## Project Context
 
-Mosaic Stack is a self-hosted, multi-user AI agent platform. TypeScript monorepo with NestJS gateway, Next.js web dashboard, Pi SDK agent runtime, and plugin architecture for Discord/Telegram.
+Mosaic Stack is a self-hosted, multi-user AI agent platform. It is a TypeScript monorepo with a NestJS gateway, Next.js dashboard, Pi SDK agent runtime, and Discord/Telegram plugin architecture.
 
-## Package Map
+### Stack
 
-| Package            | Purpose                         | Key Dependencies                 |
+- **API:** NestJS with Fastify (`apps/gateway`)
-| ------------------ | ------------------------------- | -------------------------------- |
+- **Web:** Next.js 16 with React 19 (`apps/web`)
-| `apps/gateway`     | NestJS API + WebSocket hub      | Fastify, Socket.IO, Pi SDK, OTEL |
+- **ORM and database:** Drizzle ORM, PostgreSQL 17, and pgvector (`packages/db`)
-| `apps/web`         | Next.js dashboard               | React 19, Tailwind               |
+- **Authentication:** BetterAuth (`packages/auth`)
-| `packages/types`   | Shared TypeScript contracts     | class-validator                  |
+- **Agent runtime:** Pi SDK (`apps/gateway`, `packages/mosaic`)
-| `packages/db`      | Drizzle ORM schema + migrations | drizzle-orm, postgres            |
+- **Queue:** Valkey 8 (`packages/queue`)
-| `packages/auth`    | BetterAuth configuration        | better-auth, @mosaicstack/db     |
+- **Build:** pnpm workspaces and Turborepo
-| `packages/brain`   | Data layer (PG-backed)          | @mosaicstack/db                  |
+- **CI:** Woodpecker CI
-| `packages/queue`   | Valkey task queue + MCP         | ioredis                          |
+- **Observability:** OpenTelemetry and Jaeger
-| `packages/coord`   | Mission coordination            | @mosaicstack/queue               |
-| `packages/mosaic`  | Unified `mosaic` CLI + TUI      | Ink, Pi SDK, commander           |
-| `plugins/discord`  | Discord channel plugin          | discord.js                       |
-| `plugins/telegram` | Telegram channel plugin         | Telegraf                         |
 
-## Architecture Rules
+### Package Map
 
-1. Gateway is the single API surface — all clients connect through it
+| Package            | Purpose                       | Key Dependencies                 |
-2. Pi SDK is ESM-only — gateway and CLI must use ESM
+| ------------------ | ----------------------------- | -------------------------------- |
-3. Socket.IO typed events defined in `@mosaicstack/types` enforce compile-time contracts
+| `apps/gateway`     | NestJS API + WebSocket hub    | Fastify, Socket.IO, Pi SDK, OTEL |
-4. OTEL auto-instrumentation loads before NestJS bootstrap
+| `apps/web`         | Next.js dashboard             | React 19, Tailwind               |
-5. BetterAuth manages auth tables; schema defined in `@mosaicstack/db`
+| `packages/types`   | Shared TypeScript contracts   | class-validator                  |
-6. Docker Compose provides PG (5433), Valkey (6380), OTEL Collector (4317/4318), Jaeger (16686)
+| `packages/db`      | Drizzle schema and migrations | drizzle-orm, postgres            |
-7. Explicit `@Inject()` decorators required in NestJS (tsx/esbuild doesn't emit decorator metadata)
+| `packages/auth`    | BetterAuth configuration      | better-auth, @mosaicstack/db     |
+| `packages/brain`   | Structured data layer         | @mosaicstack/db                  |
+| `packages/queue`   | Valkey task queue and MCP     | ioredis                          |
+| `packages/coord`   | Mission coordination          | @mosaicstack/queue               |
+| `packages/mosaic`  | Unified `mosaic` CLI and TUI  | Ink, Pi SDK, commander           |
+| `plugins/discord`  | Discord channel plugin        | discord.js                       |
+| `plugins/telegram` | Telegram channel plugin       | Telegraf                         |
+
+## Architecture and Code Conventions
+
+1. Gateway is the single API surface; all clients connect through it.
+2. Pi SDK is ESM-only; gateway and CLI code must remain ESM.
+3. Use `"type": "module"`, NodeNext module resolution, and `.js` extensions in imports.
+4. Keep typed Socket.IO events in `@mosaicstack/types` to enforce client/server contracts.
+5. Import OTEL tracing before NestJS bootstrap (`import './tracing.js'`).
+6. Use explicit `@Inject()` decorators in NestJS because tsx/esbuild does not emit decorator metadata.
+7. Keep DTOs in `*.dto.ts` files at module boundaries.
+8. BetterAuth owns authentication tables; their schema is defined in `@mosaicstack/db`.
+9. Create a task-specific scratchpad for non-trivial work.
 
 ## Development Workflow
 
+Requirements: Node.js 20+, pnpm 10.6.2, and Docker Compose when optional local services are needed.
+
 ```bash
-docker compose up -d          # Infrastructure
+pnpm install --frozen-lockfile
-pnpm install                  # Dependencies
+pnpm preflight
-pnpm typecheck && pnpm lint && pnpm format:check  # Quality gates
+
+# Optional local queue service only; do not start the full Compose stack.
+docker compose up -d valkey
 ```
 
-## Repo-Specific Notes
+The pre-push hook requires:
 
-- DTOs in `*.dto.ts` files at module boundaries
+```bash
-- ESM everywhere (`"type": "module"`, `.js` extensions in imports)
+pnpm preflight && pnpm typecheck && pnpm lint && pnpm format:check
-- NodeNext module resolution in all tsconfigs
+```
-- Scratchpads are mandatory for non-trivial tasks
+
+Software delivery also requires the applicable tests. Common repository commands are:
+
+```bash
+pnpm typecheck     # TypeScript checks across the workspace
+pnpm lint          # ESLint across the workspace
+pnpm test          # Checkout tests and package Vitest suites
+pnpm format:check  # Prettier check
+pnpm build         # Build all packages and applications
+```
+
+## Database and Local Runtime Safety
+
+- Current local data-layer work uses in-process PGlite; leave `DATABASE_URL` unset.
+- PostgreSQL execution is held until KBN-101-00, KBN-101-03, and KBN-101-05 land.
+- Do not invoke a migration runner, initialization SQL, or the Compose PostgreSQL service from this checkout.
+- Do not start Gateway/Web or run root `pnpm dev` as a local PGlite route. The current dotenv loader can inherit a daemon PostgreSQL DSN; KBN-101-02 must make that path fail closed first.
+- Migration artifact generation is offline and does not authorize PostgreSQL access:
+
+```bash
+pnpm --filter @mosaicstack/db db:generate
+```
 
 ## docs/TASKS.md — Schema (CANONICAL)
 

CLAUDE.md

@@ -1,46 +1,5 @@
-# CLAUDE.md — Mosaic Stack
+# Claude Compatibility Pointer
 
-## Project
+@AGENTS.md
 
-Self-hosted, multi-user AI agent platform. TypeScript monorepo.
+Do not add project guidance here. Keep `AGENTS.md` authoritative so every agent runtime receives the same instructions.
-
-## Stack
-
-- **API**: NestJS + Fastify adapter (`apps/gateway`)
-- **Web**: Next.js 16 + React 19 (`apps/web`)
-- **ORM**: Drizzle ORM + PostgreSQL 17 + pgvector (`packages/db`)
-- **Auth**: BetterAuth (`packages/auth`)
-- **Agent**: Pi SDK (`packages/agent`, `packages/mosaic`)
-- **Queue**: Valkey 8 (`packages/queue`)
-- **Build**: pnpm workspaces + Turborepo
-- **CI**: Woodpecker CI
-- **Observability**: OpenTelemetry → Jaeger
-
-## Commands
-
-```bash
-pnpm typecheck        # TypeScript check (all packages)
-pnpm lint             # ESLint (all packages)
-pnpm format:check     # Prettier check
-pnpm test             # Vitest (all packages)
-pnpm build            # Build all packages
-
-# Database
-pnpm --filter @mosaicstack/db db:generate  # Offline migration artifact generation only
-# PostgreSQL execution is held until KBN-101-00/-03/-05 land. Do not invoke a runner,
-# init SQL, or Compose PostgreSQL service from this checkout.
-
-# Dev: local PGlite data-layer work needs no PostgreSQL. Optional local queue service only:
-docker compose up -d valkey
-# Do not start Gateway/Web or root pnpm dev as a local PGlite route: the current unguarded dotenv
-# loader can inherit a daemon PostgreSQL DSN. KBN-101-02 must make that state fail closed first.
-```
-
-## Conventions
-
-- ESM everywhere (`"type": "module"`, `.js` extensions in imports)
-- NodeNext module resolution
-- Explicit `@Inject()` decorators in NestJS (tsx/esbuild doesn't support emitDecoratorMetadata)
-- DTOs in `*.dto.ts` files at module boundaries
-- OTEL tracing imported before NestJS bootstrap (`import './tracing.js'`)
-- All three gates must pass before push: typecheck, lint, format:check

README.md

@@ -48,9 +48,13 @@ mosaic wizard        # Full guided setup (gateway install → verify)
 
 ### Requirements
 
-- Node.js ≥ 20
+- Node.js ≥ 22
 - npm (for global @mosaicstack/mosaic install)
-- One or more runtimes: [Claude Code](https://docs.anthropic.com/en/docs/claude-code), [Codex](https://github.com/openai/codex), [OpenCode](https://opencode.ai), or [Pi](https://github.com/mariozechner/pi-coding-agent)
+- One or more runtimes:
+  - [Claude Code](https://docs.anthropic.com/en/docs/claude-code)
+  - [Codex](https://github.com/openai/codex)
+  - [OpenCode](https://opencode.ai)
+  - [Pi](https://pi.dev)
 
 ## Usage
 
@@ -200,7 +204,7 @@ Consent state is persisted in config. Remote upload is a no-op until you run `mo
 
 ### Prerequisites
 
-- Node.js ≥ 20
+- Node.js ≥ 22
 - pnpm 10.6+
 - Docker & Docker Compose
 

apps/gateway/src/__tests__/conversation-persistence.test.ts

@@ -417,7 +417,7 @@ describe('ConversationsController — search endpoint', () => {
       },
     ];
     brain = createMockBrain({ searchResults });
-    controller = new ConversationsController(brain as never);
+    controller = new ConversationsController(brain as never, { runtimeMode: 'legacy' });
   });
 
   it('returns matching messages for a valid search query', async () => {
@@ -479,7 +479,7 @@ describe('ConversationsController — search endpoint', () => {
 describe('ConversationsController — message CRUD', () => {
   it('listMessages returns 404 when conversation is not owned by user', async () => {
     const brain = createMockBrain({ conversation: undefined });
-    const controller = new ConversationsController(brain as never);
+    const controller = new ConversationsController(brain as never, { runtimeMode: 'legacy' });
 
     await expect(controller.listMessages(CONV_ID, { id: USER_ID })).rejects.toBeInstanceOf(
       NotFoundException,
@@ -489,7 +489,7 @@ describe('ConversationsController — message CRUD', () => {
   it('listMessages returns the messages for an owned conversation', async () => {
     const msgs = [makeMessage('user', 'Test message'), makeMessage('assistant', 'Test reply')];
     const brain = createMockBrain({ conversation: makeConversation(), messages: msgs });
-    const controller = new ConversationsController(brain as never);
+    const controller = new ConversationsController(brain as never, { runtimeMode: 'legacy' });
 
     const result = await controller.listMessages(CONV_ID, { id: USER_ID });
 
@@ -500,7 +500,7 @@ describe('ConversationsController — message CRUD', () => {
 
   it('addMessage returns the persisted message', async () => {
     const brain = createMockBrain({ conversation: makeConversation() });
-    const controller = new ConversationsController(brain as never);
+    const controller = new ConversationsController(brain as never, { runtimeMode: 'legacy' });
 
     const result = await controller.addMessage(
       CONV_ID,

apps/gateway/src/__tests__/integration/tess-cross-surface.integration.test.ts

@@ -35,6 +35,25 @@ function payload(content: string, messageId: string, correlationId: string): Dis
   };
 }
 
+/**
+ * The chat runtime router must never be exercised on the Discord approval/stop control paths —
+ * those paths run entirely through the command-authorization, runtime-provider and durable-session
+ * dependencies. Placed in the gateway's chat-runtime-router slot (the former direct `AgentService`
+ * slot) so any accidental chat-runtime dispatch throws loudly instead of silently passing. Because
+ * approval/stop never resolve a chat runtime, this fixture is never triggered and the integration
+ * stays a GREEN cross-surface control.
+ */
+function failIfUsedChatRuntimeRouter() {
+  return {
+    onModuleInit: () => {
+      throw new Error('chat runtime router must not initialise on the Discord control path');
+    },
+    get active(): never {
+      throw new Error('chat runtime must not be resolved on the Discord approval/stop path');
+    },
+  };
+}
+
 function authorization(): CommandAuthorizationService {
   const entries = new Map<string, string>();
   return new CommandAuthorizationService(
@@ -113,7 +132,7 @@ describe('interaction Discord/CLI durable-session integration', () => {
       },
     );
     const gateway = new ChatGateway(
-      {} as never,
+      failIfUsedChatRuntimeRouter() as never,
       {} as never,
       {} as never,
       {} as never,

apps/gateway/src/__tests__/resource-ownership.test.ts

@@ -60,7 +60,7 @@ describe('Resource ownership checks', () => {
     // The repo enforces ownership via the WHERE clause; it returns undefined when the
     // conversation does not belong to the requesting user.
     brain.conversations.findById.mockResolvedValue(undefined);
-    const controller = new ConversationsController(brain as never);
+    const controller = new ConversationsController(brain as never, { runtimeMode: 'legacy' });
 
     await expect(controller.findOne('conv-1', { id: 'user-1' })).rejects.toBeInstanceOf(
       NotFoundException,

apps/gateway/src/agent/__tests__/session-ownership.test.ts

@@ -1,6 +1,8 @@
+import 'reflect-metadata';
 import { readFileSync } from 'node:fs';
 import { resolve } from 'node:path';
 import { ForbiddenException, NotFoundException } from '@nestjs/common';
+import { Test, type TestingModule } from '@nestjs/testing';
 import { describe, expect, it, vi } from 'vitest';
 
 vi.mock('../agent.service.js', () => ({ AgentService: class AgentService {} }));
@@ -12,10 +14,25 @@ vi.mock('../routing/routing-engine.service.js', () => ({
 }));
 
 import { SessionsController } from '../sessions.controller.js';
+import { AgentService } from '../agent.service.js';
 import { ChatController } from '../../chat/chat.controller.js';
 import { ChatGateway } from '../../chat/chat.gateway.js';
 import type { AgentSession } from '../agent.service.js';
 import type { SessionInfoDto } from '../session.dto.js';
+import type { HarnessAdapter, HarnessConversationService } from '@mosaicstack/types';
+import { AuthGuard } from '../../auth/auth.guard.js';
+import { AUTH } from '../../auth/auth.tokens.js';
+import { BRAIN } from '../../brain/brain.tokens.js';
+import { CommandRegistryService } from '../../commands/command-registry.service.js';
+import { CommandExecutorService } from '../../commands/command-executor.service.js';
+import { RoutingEngineService } from '../routing/routing-engine.service.js';
+import { ChatRuntimeRouter } from '../../chat/chat-runtime-router.js';
+import { EmbeddedChatRuntime } from '../../chat/embedded-chat.runtime.js';
+import { ownConversation } from '../../chat/chat-runtime.js';
+import type { LegacyRuntimeStream } from '../../chat/chat-runtime.js';
+import { HarnessChatRuntime } from '../../chat/harness-chat.runtime.js';
+import { HarnessRegistry } from '../../harness/harness.registry.js';
+import { HARNESS_CONVERSATION_SERVICE_UNAVAILABLE } from '../../harness/harness.tokens.js';
 
 const USER_A = { id: 'user-a', tenantId: 'tenant-a' };
 const USER_B = { id: 'user-b', tenantId: 'tenant-b' };
@@ -74,6 +91,12 @@ function makeAgentSession(owner = USER_A): AgentSession {
   };
 }
 
+/**
+ * A shape-complete, non-throwing AgentService fake scoped so that USER_B (a foreign owner guessing
+ * USER_A's conversation id) is never granted the session. Because every method exists and no method
+ * throws for a wrong shape, production runs to its real ownership decision — the RED never comes from
+ * a `getSession is not a function` TypeError, only from a router-boundary/scope assertion mismatch.
+ */
 function makeScopedAgentService() {
   const foreign = makeAgentSession(USER_A);
   return {
@@ -87,7 +110,7 @@ function makeScopedAgentService() {
     getSession: vi.fn((_id: string, scope?: { userId: string; tenantId?: string }) =>
       scope?.userId === USER_B.id ? undefined : foreign,
     ),
-    createSession: vi.fn().mockRejectedValue(new ForbiddenException('Session scope mismatch')),
+    createSession: vi.fn().mockRejectedValue(new NotFoundException('Session scope mismatch')),
     onEvent: vi.fn(() => vi.fn()),
     addChannel: vi.fn(),
     removeChannel: vi.fn(),
@@ -96,6 +119,201 @@ function makeScopedAgentService() {
   };
 }
 
+type ScopedAgentService = ReturnType<typeof makeScopedAgentService>;
+
+/**
+ * A structurally-complete harness conversation service that throws if any method is invoked.
+ * Fronted behind the legacy runtime's harness slot: the legacy path must never reach it.
+ */
+const failIfUsedConversationService = {
+  attach: () => {
+    throw new Error('harness conversation service must not be reached on the legacy path');
+  },
+  detach: () => {
+    throw new Error('harness conversation service must not be reached on the legacy path');
+  },
+  send: () => {
+    throw new Error('harness conversation service must not be reached on the legacy path');
+  },
+
+  subscribeFrom: async function* () {
+    throw new Error('harness conversation service must not be reached on the legacy path');
+  },
+} as unknown as HarnessConversationService;
+
+/** A structurally-complete, non-sentinel conversation service used to satisfy the pi-rpc readiness gate. */
+const boundConversationService = {
+  attach: () => Promise.reject(new Error('unused')),
+  detach: () => Promise.reject(new Error('unused')),
+  send: () => Promise.reject(new Error('unused')),
+
+  subscribeFrom: async function* () {
+    throw new Error('unused');
+  },
+} as unknown as HarnessConversationService;
+
+function registryWith(adapterIds: readonly string[]): HarnessRegistry {
+  const registry = new HarnessRegistry();
+  for (const id of adapterIds) {
+    registry.register({
+      id,
+      describe: () => Promise.reject(new Error('unused')),
+      catalog: () => Promise.reject(new Error('unused')),
+      create: () => Promise.reject(new Error('unused')),
+      resume: () => Promise.reject(new Error('unused')),
+    } as HarnessAdapter);
+  }
+  return registry;
+}
+
+/**
+ * Build the real legacy-mode {@link ChatRuntimeRouter} fronting a real {@link EmbeddedChatRuntime}
+ * that holds the scoped AgentService fake. This is the ONLY path server-derived scope may travel to
+ * reach an AgentService: controller/gateway → ChatRuntimeRouter → EmbeddedChatRuntime → AgentService.
+ * The `embeddedAgentService` handed here is a SEPARATE instance from the directly-injected fake, so a
+ * call landing on it proves the router-delegation redesign is live rather than the old direct path.
+ */
+function legacyRouterFronting(agentService: unknown): ChatRuntimeRouter {
+  const embedded = new EmbeddedChatRuntime(agentService as never);
+  const harness = new HarnessChatRuntime(failIfUsedConversationService);
+  const router = new ChatRuntimeRouter(
+    new HarnessRegistry(),
+    HARNESS_CONVERSATION_SERVICE_UNAVAILABLE,
+    embedded,
+    harness,
+    'legacy',
+  );
+  router.onModuleInit();
+  return router;
+}
+
+/**
+ * The AgentService method names the controller/gateway must NEVER drive on the runtime at the
+ * delegation boundary. An AgentService-shaped router shim (a method-for-method mirror) would record
+ * one of these instead of the frozen legacy op, so asserting their ABSENCE from the observed runtime
+ * call set defeats the shim on INVOCATION evidence — never satisfiable by dead source text.
+ */
+const FORBIDDEN_AGENT_OPS = [
+  'getSession',
+  'createSession',
+  'onEvent',
+  'addChannel',
+  'prompt',
+  'setThinking',
+  'abort',
+] as const;
+
+/**
+ * Wrap a real {@link ChatRuntimeRouter} in a call-recording Proxy. Every property access that yields
+ * an OWN/inherited callable is returned as a thin wrapper that appends the method name to `calls` at
+ * INVOCATION time and forwards to the real method (bound to the real target, so the router's internal
+ * delegation to the embedded runtime runs untouched below this boundary). Non-function and MISSING
+ * properties are returned verbatim via Reflect.get — the observer NEVER fabricates a value, returns a
+ * canned outcome, or delegates a not-yet-implemented named op, so it cannot itself become a shim.
+ *
+ * The result is a RUNTIME call set of exactly the methods the controller/gateway invoke ON the router
+ * at the delegation seam. Only an actual call can enter it; a dead method, comment, or string in the
+ * production source cannot. This replaces the earlier `source.toContain('<frozen op>')` proof — which
+ * a dead declaration could satisfy while production still executed a shim — with invocation evidence.
+ */
+function makeRecordingRouter(target: ChatRuntimeRouter, calls: string[]): ChatRuntimeRouter {
+  return new Proxy(target, {
+    get(t, prop) {
+      const value = Reflect.get(t, prop);
+      if (typeof value === 'function' && typeof prop === 'string') {
+        return (...args: unknown[]) => {
+          calls.push(prop);
+          return (value as (...a: unknown[]) => unknown).apply(t, args);
+        };
+      }
+      return value;
+    },
+  }) as ChatRuntimeRouter;
+}
+
+/**
+ * Real Nest DI dual-provider fixture (mirrors the blessed group-3 pattern in chat-security.test.ts).
+ *
+ * BOTH an `AgentService` provider (the FORBIDDEN direct dependency) and a `ChatRuntimeRouter` provider
+ * (fronting a real EmbeddedChatRuntime over a SEPARATE scoped AgentService) are registered. Production
+ * resolves whichever its constructor declares:
+ *   - RED today: the controller/gateway `@Inject(AgentService)` → the direct fake is consulted, the
+ *     router (and its embedded fake) is never reached.
+ *   - GREEN later: the controller/gateway inject `ChatRuntimeRouter` → the direct fake is never
+ *     touched (stays at zero) and scope is observed inside the embedded fake behind the router.
+ * The SAME test body reds today and greens later; a method-for-method AgentService shim on the router
+ * records a FORBIDDEN op (and never the frozen legacy op) in the observed runtime call set, and
+ * restoring the direct injection cannot satisfy the "direct fake at zero" / "embedded fake observed
+ * scope" / "frozen op invoked on the router" anchors. The router is wrapped by {@link
+ * makeRecordingRouter} so those anchors are runtime invocation evidence, not source substrings.
+ */
+function buildRestModule(
+  directAgentService: ScopedAgentService,
+  embeddedAgentService: ScopedAgentService,
+  routerCalls: string[],
+): Promise<TestingModule> {
+  return (
+    Test.createTestingModule({
+      controllers: [ChatController],
+      providers: [
+        { provide: AgentService, useValue: directAgentService },
+        {
+          provide: ChatRuntimeRouter,
+          useFactory: () =>
+            makeRecordingRouter(legacyRouterFronting(embeddedAgentService), routerCalls),
+        },
+      ],
+    })
+      // ChatController's @UseGuards(AuthGuard) is resolved during instance loading; AuthGuard injects
+      // AUTH, an HTTP-only concern never exercised by a direct handler call. Stub it so the graph
+      // resolves and the test reds on BEHAVIOUR, not on a DI collection error.
+      .overrideGuard(AuthGuard)
+      .useValue({ canActivate: () => true })
+      .compile()
+  );
+}
+
+function buildGatewayModule(
+  directAgentService: ScopedAgentService,
+  embeddedAgentService: ScopedAgentService,
+  routerCalls: string[],
+): Promise<TestingModule> {
+  const brain = {
+    conversations: {
+      // The sender OWNS this durable conversation, so the browser-send admission gate lets the turn
+      // reach the router seam. Foreignness is asserted downstream at the in-memory agent session
+      // (getSession({USER_B}) -> undefined), not at durable admission — the admission-rejection
+      // property has its own dedicated coverage.
+      findById: vi.fn().mockResolvedValue({ id: CONVERSATION_ID, userId: USER_B.id }),
+      create: vi.fn().mockResolvedValue(undefined),
+      update: vi.fn().mockResolvedValue(undefined),
+      findMessages: vi.fn().mockResolvedValue([]),
+      addMessage: vi.fn().mockResolvedValue({ id: 'persisted-turn' }),
+    },
+  };
+  return Test.createTestingModule({
+    providers: [
+      ChatGateway,
+      { provide: AgentService, useValue: directAgentService },
+      { provide: AUTH, useValue: { api: { getSession: vi.fn().mockResolvedValue(null) } } },
+      { provide: BRAIN, useValue: brain },
+      { provide: CommandRegistryService, useValue: { getManifest: vi.fn().mockReturnValue([]) } },
+      { provide: CommandExecutorService, useValue: { execute: vi.fn() } },
+      {
+        provide: RoutingEngineService,
+        useValue: {
+          resolve: vi.fn().mockResolvedValue({ provider: 'test', model: 'test-model' }),
+        },
+      },
+      {
+        provide: ChatRuntimeRouter,
+        useFactory: () =>
+          makeRecordingRouter(legacyRouterFronting(embeddedAgentService), routerCalls),
+      },
+    ],
+  }).compile();
+}
+
 describe('TESS-M1-SEC-002 AgentService ownership boundary', () => {
   it('requires explicit owner+tenant scope on protected session operations', () => {
     const source = readFileSync(resolve('src/agent/agent.service.ts'), 'utf8');
@@ -152,50 +370,66 @@ describe('TESS-M1-SEC-002 REST session ownership and tenant binding', () => {
   });
 });
 
-describe('TESS-M1-SEC-002 REST chat send ownership and tenant binding', () => {
+describe('TESS-M1-SEC-002 REST chat send ownership and tenant binding (router-delegated legacy runtime)', () => {
-  it('does not send a prompt into another owner/tenant session by guessed conversationId', async () => {
+  // TESS test A — REST /api/chat send. The genuine RED is the router-delegation redesign, not a slot
-    const agentService = makeScopedAgentService();
+  // swap: the forbidden directly-injected AgentService must go UNtouched while the server-derived
-    const controller = new ChatController(agentService as never);
+  // scope is observed inside the real ChatRuntimeRouter → EmbeddedChatRuntime → AgentService path.
+  it('routes a REST send through completeLegacyRestTurn and never the directly-injected AgentService', async () => {
+    const directAgentService = makeScopedAgentService(); // FORBIDDEN direct dependency
+    const embeddedAgentService = makeScopedAgentService(); // reached ONLY via router → embedded delegation
+    const routerCalls: string[] = []; // runtime call set observed AT the controller → router seam
+    const moduleRef = await buildRestModule(directAgentService, embeddedAgentService, routerCalls);
+    try {
+      const controller = moduleRef.get(ChatController, { strict: false });
 
-    await expect(
+      // Foreign ownership is denied (never resolves) — a control that holds today AND at GREEN.
-      controller.chat({ conversationId: CONVERSATION_ID, content: 'take over' }, USER_B),
+      await expect(
-    ).rejects.toMatchObject({ status: 404 });
+        controller.chat({ conversationId: CONVERSATION_ID, content: 'take over' }, USER_B),
+      ).rejects.toBeDefined();
 
-    expect(agentService.getSession).toHaveBeenCalledWith(CONVERSATION_ID, {
+      // Soft anchors so EVERY anchor is evaluated under each mutation, not just the first to fail.
-      userId: USER_B.id,
+
-      tenantId: USER_B.tenantId,
+      // RUNTIME anchor A1 — delegation: the controller must INVOKE the frozen legacy op on the router.
-    });
+      // Only an actual call enters routerCalls; a dead method/comment/string cannot. RED today (the
-    expect(agentService.prompt).not.toHaveBeenCalled();
+      // controller @Inject(AgentService) and never calls the router). GREEN once it drives the op.
+      expect
+        .soft(routerCalls, 'controller must invoke completeLegacyRestTurn on the router')
+        .toContain('completeLegacyRestTurn');
+      // RUNTIME anchor A2 — nondelegation: the controller must not drive any AgentService-shaped op on
+      // the router. An AgentService-shaped router shim records one of these → RED, defeating the shim
+      // on invocation evidence (not source text). A dead named method added alongside the shim does not
+      // help: it is never invoked, so it never enters routerCalls while a forbidden op still does.
+      for (const op of FORBIDDEN_AGENT_OPS) {
+        expect
+          .soft(routerCalls, `router seam must not invoke AgentService.${op}`)
+          .not.toContain(op);
+      }
+      // RUNTIME anchor A3 — the forbidden directly-injected AgentService stays at zero (fails today;
+      // restoring the direct injection keeps it failing).
+      expect.soft(directAgentService.getSession).not.toHaveBeenCalled();
+      // RUNTIME anchor A4 — server-derived scope observed INSIDE the separate embedded fake behind the
+      // router (fails today; the router path is never taken).
+      expect.soft(embeddedAgentService.getSession).toHaveBeenCalledWith(CONVERSATION_ID, {
+        userId: USER_B.id,
+        tenantId: USER_B.tenantId,
+      });
+
+      // Zero foreign mutation on either path (holds today and at GREEN).
+      expect.soft(directAgentService.prompt).not.toHaveBeenCalled();
+      expect.soft(embeddedAgentService.prompt).not.toHaveBeenCalled();
+
+      // Defense-in-depth (NOT load-bearing; the runtime anchors above carry the anti-mask): the
+      // controller no longer declares the direct embedded AgentService dependency. A negative source
+      // check cannot be satisfied by dead text — it only fails when the injection is present.
+      const controllerSource = readFileSync(resolve('src/chat/chat.controller.ts'), 'utf8');
+      expect.soft(controllerSource).not.toContain('@Inject(AgentService)');
+    } finally {
+      await moduleRef.close();
+    }
   });
 });
 
-describe('TESS-M1-SEC-002 WebSocket session ownership and tenant binding', () => {
+describe('TESS-M1-SEC-002 WebSocket session ownership and tenant binding (router-delegated legacy runtime)', () => {
-  function makeGateway(agentService = makeScopedAgentService()) {
-    const brain = {
-      conversations: {
-        findById: vi.fn().mockResolvedValue(undefined),
-        create: vi.fn().mockResolvedValue(undefined),
-        update: vi.fn().mockResolvedValue(undefined),
-        findMessages: vi.fn().mockResolvedValue([]),
-        addMessage: vi.fn().mockResolvedValue(undefined),
-      },
-    };
-    const commandRegistry = { getManifest: vi.fn().mockReturnValue([]) };
-    const commandExecutor = { execute: vi.fn() };
-    const routingEngine = {
-      resolve: vi.fn().mockResolvedValue({ provider: 'test', model: 'test-model' }),
-    };
-    const gateway = new ChatGateway(
-      agentService as never,
-      {} as never,
-      brain as never,
-      commandRegistry as never,
-      commandExecutor as never,
-      routingEngine as never,
-    );
-    return { gateway, agentService };
-  }
-
   function makeSocket() {
     return {
       id: 'socket-b',
@@ -206,57 +440,519 @@ describe('TESS-M1-SEC-002 WebSocket session ownership and tenant binding', () =>
     };
   }
 
-  it('does not attach or send to another owner/tenant session by guessed conversationId', async () => {
+  // TESS test B — WebSocket send/attach.
-    const { gateway, agentService } = makeGateway();
+  it('routes a WebSocket send through prepareLegacySocketTurn and never the directly-injected AgentService', async () => {
-    const socket = makeSocket();
+    const directAgentService = makeScopedAgentService();
+    const embeddedAgentService = makeScopedAgentService();
+    const routerCalls: string[] = [];
+    const moduleRef = await buildGatewayModule(
+      directAgentService,
+      embeddedAgentService,
+      routerCalls,
+    );
+    try {
+      const gateway = moduleRef.get(ChatGateway, { strict: false });
+      const socket = makeSocket();
 
-    await gateway.handleMessage(socket as never, {
+      await Promise.resolve(
-      conversationId: CONVERSATION_ID,
+        gateway.handleMessage(socket as never, {
-      content: 'attach to foreign session',
+          conversationId: CONVERSATION_ID,
-    });
+          content: 'attach to foreign session',
+        }),
+      ).catch(() => undefined);
 
-    expect(agentService.getSession).toHaveBeenCalledWith(CONVERSATION_ID, {
+      // RUNTIME anchor B1 — delegation: the gateway must invoke the frozen socket op on the router.
-      userId: USER_B.id,
+      expect
-      tenantId: USER_B.tenantId,
+        .soft(routerCalls, 'gateway must invoke prepareLegacySocketTurn on the router')
-    });
+        .toContain('prepareLegacySocketTurn');
-    expect(agentService.onEvent).not.toHaveBeenCalled();
+      // RUNTIME anchor B2 — nondelegation: no AgentService-shaped op on the router (defeats the shim).
-    expect(agentService.addChannel).not.toHaveBeenCalled();
+      for (const op of FORBIDDEN_AGENT_OPS) {
+        expect
+          .soft(routerCalls, `router seam must not invoke AgentService.${op}`)
+          .not.toContain(op);
+      }
+      // RED anchor B3 — forbidden direct AgentService untouched (fails today, gateway injects it).
+      expect.soft(directAgentService.getSession).not.toHaveBeenCalled();
+      // RED anchor B4 — scope observed inside router → embedded delegation (fails today, never reached).
+      expect.soft(embeddedAgentService.getSession).toHaveBeenCalledWith(CONVERSATION_ID, {
+        userId: USER_B.id,
+        tenantId: USER_B.tenantId,
+      });
+      // Foreign session gets zero lease/listener/channel/prompt on EITHER path (holds today and GREEN).
+      expect.soft(directAgentService.onEvent).not.toHaveBeenCalled();
+      expect.soft(directAgentService.addChannel).not.toHaveBeenCalled();
+      expect.soft(directAgentService.prompt).not.toHaveBeenCalled();
+      expect.soft(embeddedAgentService.onEvent).not.toHaveBeenCalled();
+      expect.soft(embeddedAgentService.addChannel).not.toHaveBeenCalled();
+      expect.soft(embeddedAgentService.prompt).not.toHaveBeenCalled();
+      expect
+        .soft(socket.emit)
+        .toHaveBeenCalledWith(
+          'error',
+          expect.objectContaining({ conversationId: CONVERSATION_ID }),
+        );
+
+      // Defense-in-depth (NOT load-bearing): gateway no longer declares the direct dependency.
+      const gatewaySource = readFileSync(resolve('src/chat/chat.gateway.ts'), 'utf8');
+      expect.soft(gatewaySource).not.toContain('@Inject(AgentService)');
+    } finally {
+      await moduleRef.close();
+    }
+  });
+
+  // TESS test C — WebSocket set:thinking.
+  it('routes set:thinking through setLegacyThinking and never the directly-injected AgentService', async () => {
+    const directAgentService = makeScopedAgentService();
+    const embeddedAgentService = makeScopedAgentService();
+    const routerCalls: string[] = [];
+    const moduleRef = await buildGatewayModule(
+      directAgentService,
+      embeddedAgentService,
+      routerCalls,
+    );
+    try {
+      const gateway = moduleRef.get(ChatGateway, { strict: false });
+      const socket = makeSocket();
+
+      await Promise.resolve(
+        gateway.handleSetThinking(socket as never, {
+          conversationId: CONVERSATION_ID,
+          level: 'high',
+        }),
+      ).catch(() => undefined);
+
+      // RUNTIME anchor C1 — delegation: the gateway must invoke the frozen thinking op on the router.
+      expect
+        .soft(routerCalls, 'gateway must invoke setLegacyThinking on the router')
+        .toContain('setLegacyThinking');
+      // RUNTIME anchor C2 — nondelegation: no AgentService-shaped op on the router (defeats the shim).
+      for (const op of FORBIDDEN_AGENT_OPS) {
+        expect
+          .soft(routerCalls, `router seam must not invoke AgentService.${op}`)
+          .not.toContain(op);
+      }
+      expect.soft(directAgentService.getSession).not.toHaveBeenCalled();
+      expect.soft(embeddedAgentService.getSession).toHaveBeenCalledWith(CONVERSATION_ID, {
+        userId: USER_B.id,
+        tenantId: USER_B.tenantId,
+      });
+      expect
+        .soft(socket.emit)
+        .toHaveBeenCalledWith(
+          'error',
+          expect.objectContaining({ conversationId: CONVERSATION_ID }),
+        );
+    } finally {
+      await moduleRef.close();
+    }
+  });
+
+  // TESS test D — WebSocket abort.
+  it('routes abort through abortLegacyTurn and never the directly-injected AgentService', async () => {
+    const directAgentService = makeScopedAgentService();
+    const embeddedAgentService = makeScopedAgentService();
+    const routerCalls: string[] = [];
+    const moduleRef = await buildGatewayModule(
+      directAgentService,
+      embeddedAgentService,
+      routerCalls,
+    );
+    try {
+      const gateway = moduleRef.get(ChatGateway, { strict: false });
+      const socket = makeSocket();
+
+      await Promise.resolve(
+        gateway.handleAbort(socket as never, { conversationId: CONVERSATION_ID }),
+      ).catch(() => undefined);
+
+      // RUNTIME anchor D1 — delegation: the gateway must invoke the frozen abort op on the router.
+      expect
+        .soft(routerCalls, 'gateway must invoke abortLegacyTurn on the router')
+        .toContain('abortLegacyTurn');
+      // RUNTIME anchor D2 — nondelegation: no AgentService-shaped op on the router (defeats the shim).
+      for (const op of FORBIDDEN_AGENT_OPS) {
+        expect
+          .soft(routerCalls, `router seam must not invoke AgentService.${op}`)
+          .not.toContain(op);
+      }
+      expect.soft(directAgentService.getSession).not.toHaveBeenCalled();
+      expect.soft(embeddedAgentService.getSession).toHaveBeenCalledWith(CONVERSATION_ID, {
+        userId: USER_B.id,
+        tenantId: USER_B.tenantId,
+      });
+      expect
+        .soft(socket.emit)
+        .toHaveBeenCalledWith(
+          'error',
+          expect.objectContaining({ conversationId: CONVERSATION_ID }),
+        );
+    } finally {
+      await moduleRef.close();
+    }
+  });
+
+  // TESS test E (genuine, unchanged) — pi-rpc browser-legacy refusal.
+  it('rejects a browser legacy raw message in pi-rpc mode with a fixed typed unsupported and executes nothing', async () => {
+    // pi-rpc: the harness runtime is live. The browser legacy `message` path is unsupported and
+    // must be refused with a fixed typed code, touching neither the embedded AgentService nor the
+    // harness conversation service.
+    const agentService = makeScopedAgentService();
+    const embedded = new EmbeddedChatRuntime(agentService as never);
+    const harnessConversation = {
+      attach: vi.fn(),
+      detach: vi.fn(),
+      send: vi.fn(),
+      subscribeFrom: vi.fn(),
+    };
+    const harness = new HarnessChatRuntime(harnessConversation as never);
+    const router = new ChatRuntimeRouter(
+      registryWith(['pi']),
+      boundConversationService,
+      embedded,
+      harness,
+      'pi-rpc',
+    );
+    router.onModuleInit();
+
+    const brain = {
+      conversations: {
+        findById: vi.fn().mockResolvedValue(undefined),
+        create: vi.fn().mockResolvedValue(undefined),
+        update: vi.fn().mockResolvedValue(undefined),
+        findMessages: vi.fn().mockResolvedValue([]),
+        addMessage: vi.fn().mockResolvedValue(undefined),
+      },
+    };
+    const gateway = new ChatGateway(
+      router as never,
+      {} as never,
+      brain as never,
+      { getManifest: vi.fn().mockReturnValue([]) } as never,
+      { execute: vi.fn() } as never,
+      { resolve: vi.fn() } as never,
+    );
+    const socket = {
+      id: 'socket-b',
+      connected: true,
+      data: { user: USER_B, session: { id: 'auth-session-b', userId: USER_B.id } },
+      emit: vi.fn(),
+      disconnect: vi.fn(),
+    };
+
+    await Promise.resolve(
+      gateway.handleMessage(socket as never, {
+        conversationId: CONVERSATION_ID,
+        content: 'route me',
+      }),
+    ).catch(() => undefined);
+
+    expect(socket.emit).toHaveBeenCalledWith(
+      'error',
+      expect.objectContaining({ code: 'runtime_unsupported' }),
+    );
+    expect(agentService.getSession).not.toHaveBeenCalled();
     expect(agentService.prompt).not.toHaveBeenCalled();
-    expect(socket.emit).toHaveBeenCalledWith(
+    expect(harnessConversation.attach).not.toHaveBeenCalled();
-      'error',
+    expect(harnessConversation.send).not.toHaveBeenCalled();
-      expect.objectContaining({ conversationId: CONVERSATION_ID }),
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Task-5 AMEND — embedded runtime lease lifecycle (G1) + ownership collapse (G5).
+// These drive the real EmbeddedChatRuntime directly over a shape-complete AgentService
+// fake (every touched method exists, so a RED can only come from behavior, never a
+// `getSession is not a function` TypeError). Ownership context is minted through the
+// real `ownConversation` factory — the only sanctioned way to reach a port op.
+// ---------------------------------------------------------------------------
+
+const EMBEDDED_SCOPE = { userId: USER_A.id, tenantId: USER_A.tenantId };
+const CONVERSATION_UNAVAILABLE_RESULT = {
+  ok: false,
+  code: 'conversation_unavailable',
+  retryable: false,
+} as const;
+
+/** A stream sink; `channelId` is server-derived, `onEvent` records nothing here. */
+function makeStream(): LegacyRuntimeStream {
+  return { channelId: 'websocket:test-1', onEvent: vi.fn() };
+}
+
+/**
+ * getSession → undefined (session missing), createSession → rejects with `err`. Exercises the
+ * `resolveOrCreate` collapse branch. `prompt` exists so its ABSENCE from the call record proves
+ * the turn short-circuited before any dispatch.
+ */
+function makeCollapsingAgentService(err: Error) {
+  return {
+    getSession: vi.fn(() => undefined),
+    createSession: vi.fn().mockRejectedValue(err),
+    onEvent: vi.fn(() => vi.fn()),
+    addChannel: vi.fn(),
+    removeChannel: vi.fn(),
+    prompt: vi.fn().mockResolvedValue(undefined),
+    recordTokenUsage: vi.fn(),
+  };
+}
+
+/** getSession → a live owned session, so `resolveOrCreate` succeeds and a lease is built. */
+function makeLeaseAgentService() {
+  const session = makeAgentSession(USER_A);
+  const unsubscribe = vi.fn();
+  const svc = {
+    getSession: vi.fn(() => session),
+    createSession: vi.fn(),
+    onEvent: vi.fn(() => unsubscribe),
+    addChannel: vi.fn(),
+    removeChannel: vi.fn(),
+    prompt: vi.fn().mockResolvedValue(undefined),
+    recordTokenUsage: vi.fn(),
+  };
+  return { svc, unsubscribe, session };
+}
+
+/**
+ * getSession → a live owned session (REST resolveOrCreate succeeds), onEvent returns a `detach`
+ * spy, and `prompt` REJECTS with a non-timeout error. Drives the REST-turn catch path so the single
+ * idempotent teardown must clear the 120s timeout and detach the listener exactly once.
+ */
+function makeRejectingPromptAgentService() {
+  const session = makeAgentSession(USER_A);
+  const detach = vi.fn();
+  const svc = {
+    getSession: vi.fn(() => session),
+    createSession: vi.fn(),
+    onEvent: vi.fn(() => detach),
+    addChannel: vi.fn(),
+    removeChannel: vi.fn(),
+    prompt: vi.fn().mockRejectedValue(new Error('agent backend exploded')),
+    recordTokenUsage: vi.fn(),
+  };
+  return { svc, detach };
+}
+
+describe('TESS Task-5 embedded ownership collapse (missing and foreign are indistinguishable, never throw)', () => {
+  const ctx = ownConversation(CONVERSATION_ID, EMBEDDED_SCOPE);
+
+  it('collapses a foreign (Forbidden) create to conversation_unavailable and never throws', async () => {
+    const svc = makeCollapsingAgentService(new ForbiddenException('foreign owner'));
+    const runtime = new EmbeddedChatRuntime(svc as never);
+
+    const result = await runtime.completeLegacyRestTurn(ctx, { content: 'take over' });
+
+    expect(result).toEqual(CONVERSATION_UNAVAILABLE_RESULT);
+    expect(svc.prompt).not.toHaveBeenCalled();
+  });
+
+  it('collapses a missing (NotFound) create to conversation_unavailable and never throws', async () => {
+    const svc = makeCollapsingAgentService(new NotFoundException('no such conversation'));
+    const runtime = new EmbeddedChatRuntime(svc as never);
+
+    const result = await runtime.completeLegacyRestTurn(ctx, { content: 'hello' });
+
+    expect(result).toEqual(CONVERSATION_UNAVAILABLE_RESULT);
+    expect(svc.prompt).not.toHaveBeenCalled();
+  });
+
+  it('returns the IDENTICAL collapse for foreign and missing so neither can be distinguished', async () => {
+    const foreign = new EmbeddedChatRuntime(
+      makeCollapsingAgentService(new ForbiddenException('foreign owner')) as never,
     );
+    const missing = new EmbeddedChatRuntime(
+      makeCollapsingAgentService(new NotFoundException('no such conversation')) as never,
+    );
+
+    const foreignResult = await foreign.completeLegacyRestTurn(ctx, { content: 'x' });
+    const missingResult = await missing.completeLegacyRestTurn(ctx, { content: 'x' });
+
+    expect(foreignResult).toEqual(missingResult);
+    expect(foreignResult).toEqual(CONVERSATION_UNAVAILABLE_RESULT);
   });
+});
+
+describe('TESS Task-5 embedded socket lease lifecycle (one-shot dispatch, idempotent dispose, partial-setup rollback)', () => {
+  const ctx = ownConversation(CONVERSATION_ID, EMBEDDED_SCOPE);
+
+  it('dispatches the turn exactly once; a second dispatch is a no-op turn_already_dispatched', async () => {
+    const { svc } = makeLeaseAgentService();
+    const runtime = new EmbeddedChatRuntime(svc as never);
 
-  it('does not mutate thinking level on another owner/tenant session', () => {
+    const prepared = await runtime.prepareLegacySocketTurn(ctx, { content: 'first' }, makeStream());
-    const { gateway, agentService } = makeGateway();
+    expect(prepared.ok).toBe(true);
-    const socket = makeSocket();
+    if (!prepared.ok) throw new Error('prepareLegacySocketTurn should succeed');
+    const lease = prepared.value;
 
-    gateway.handleSetThinking(socket as never, { conversationId: CONVERSATION_ID, level: 'high' });
+    const first = await lease.dispatch();
+    expect(first).toEqual({ ok: true, value: undefined });
+    expect(svc.prompt).toHaveBeenCalledTimes(1);
 
-    expect(agentService.getSession).toHaveBeenCalledWith(CONVERSATION_ID, {
+    const second = await lease.dispatch();
-      userId: USER_B.id,
+    expect(second).toEqual({ ok: false, code: 'turn_already_dispatched', retryable: false });
-      tenantId: USER_B.tenantId,
+    // Zero additional effect — the second dispatch must not prompt again.
+    expect(svc.prompt).toHaveBeenCalledTimes(1);
+  });
+
+  it('disposes once; a second dispose is a silent no-op that never re-detaches or destroys the session', async () => {
+    const { svc, unsubscribe, session } = makeLeaseAgentService();
+    const runtime = new EmbeddedChatRuntime(svc as never);
+
+    const prepared = await runtime.prepareLegacySocketTurn(ctx, { content: 'x' }, makeStream());
+    expect(prepared.ok).toBe(true);
+    if (!prepared.ok) throw new Error('prepareLegacySocketTurn should succeed');
+    const lease = prepared.value;
+
+    await lease.dispose();
+    await lease.dispose();
+
+    // Listener + channel torn down exactly once across two dispose calls.
+    expect(unsubscribe).toHaveBeenCalledTimes(1);
+    expect(svc.removeChannel).toHaveBeenCalledTimes(1);
+    // Disposal never terminates the underlying session or process.
+    expect(session.piSession.abort).not.toHaveBeenCalled();
+    expect(session.piSession.dispose).not.toHaveBeenCalled();
+  });
+
+  it('rolls back the acquired listener and returns a total safe failure when channel attach fails mid-setup', async () => {
+    const { svc, unsubscribe } = makeLeaseAgentService();
+    svc.addChannel = vi.fn(() => {
+      throw new Error('channel attach failed');
     });
-    expect(socket.emit).toHaveBeenCalledWith(
+    const runtime = new EmbeddedChatRuntime(svc as never);
-      'error',
+
-      expect.objectContaining({ conversationId: CONVERSATION_ID }),
+    // Must NOT throw out of the port — a partial setup collapses to a total safe failure.
-    );
+    const prepared = await runtime.prepareLegacySocketTurn(ctx, { content: 'x' }, makeStream());
+    expect(prepared.ok).toBe(false);
+    // Exactly what was acquired (the event listener) is rolled back.
+    expect(unsubscribe).toHaveBeenCalledTimes(1);
+  });
+});
+
+describe('TESS Task-5 embedded REST turn teardown (a prompt rejection frees the timer + listener exactly once)', () => {
+  const ctx = ownConversation(CONVERSATION_ID, EMBEDDED_SCOPE);
+
+  it('clears the 120s timeout and detaches the listener exactly once when prompt() rejects, leaving no timer to reject the abandoned done-promise later (Task 5 finding 6)', async () => {
+    const { svc, detach } = makeRejectingPromptAgentService();
+    const runtime = new EmbeddedChatRuntime(svc as never);
+
+    // A rejected `done` promise firing after completeLegacyRestTurn has already returned would
+    // surface as an unhandledRejection — the leak this test fences. Capture any that escape.
+    const unhandled: unknown[] = [];
+    const onUnhandled = (reason: unknown): void => {
+      unhandled.push(reason);
+    };
+    process.on('unhandledRejection', onUnhandled);
+    vi.useFakeTimers();
+    try {
+      const result = await runtime.completeLegacyRestTurn(ctx, {
+        content: 'trigger a backend failure',
+      });
+
+      // The rejection collapses to a total safe failure (not a timeout) — never throws out of the port.
+      expect(result).toEqual({ ok: false, code: 'operation_failed', retryable: false });
+      // The single idempotent dispose ran in the catch: listener detached exactly once.
+      expect(detach).toHaveBeenCalledTimes(1);
+
+      // dispose() cleared the REST timeout, so advancing far past it (120s) fires nothing: no second
+      // detach, and — the actual leak — no live timer left to reject the now-abandoned `done` promise.
+      vi.advanceTimersByTime(600_000);
+      expect(detach).toHaveBeenCalledTimes(1);
+    } finally {
+      vi.useRealTimers();
+    }
+    // Let any scheduled rejection surface on a real macrotask, then confirm none did.
+    await new Promise((resolve) => setTimeout(resolve, 0));
+    process.off('unhandledRejection', onUnhandled);
+    expect(unhandled).toHaveLength(0);
   });
+
+  it('bounds a hung prompt: when prompt() never settles and no agent_end arrives, the 120s timeout ends the turn with a timeout result and exactly one teardown, no unhandledRejection (Task 5 finding 6 — pending-prompt timeout)', async () => {
+    const session = makeAgentSession(USER_A);
+    const detach = vi.fn();
+    const svc = {
+      getSession: vi.fn(() => session),
+      createSession: vi.fn(),
+      onEvent: vi.fn(() => detach),
+      addChannel: vi.fn(),
+      removeChannel: vi.fn(),
+      // The prompt never resolves or rejects — a hung agent backend. Under the pre-fix sequential
+      // `await prompt()` the timer could never even be observed, so the turn hung forever.
+      prompt: vi.fn(() => new Promise<void>(() => undefined)),
+      recordTokenUsage: vi.fn(),
+    };
+    const runtime = new EmbeddedChatRuntime(svc as never);
 
-  it('does not terminate another owner/tenant session over WebSocket abort', async () => {
+    const unhandled: unknown[] = [];
-    const { gateway, agentService } = makeGateway();
+    const onUnhandled = (reason: unknown): void => {
-    const socket = makeSocket();
+      unhandled.push(reason);
+    };
+    process.on('unhandledRejection', onUnhandled);
+    vi.useFakeTimers();
+    try {
+      const resultPromise = runtime.completeLegacyRestTurn(ctx, {
+        content: 'a prompt that never returns',
+      });
+      // No agent_end, prompt still pending: only the 120s timeout can end the turn. Promise.all
+      // installed a handler on `done` synchronously, so the timer bounds the turn while prompt hangs.
+      await vi.advanceTimersByTimeAsync(200_000);
+      const result = await resultPromise;
 
-    await gateway.handleAbort(socket as never, { conversationId: CONVERSATION_ID });
+      expect(result).toEqual({ ok: false, code: 'timeout', retryable: true });
+      // The single idempotent dispose ran on the timeout path: listener detached exactly once.
+      expect(detach).toHaveBeenCalledTimes(1);
+      // Advancing far past the deadline fires nothing more: dispose cleared the timer.
+      vi.advanceTimersByTime(600_000);
+      expect(detach).toHaveBeenCalledTimes(1);
+    } finally {
+      vi.useRealTimers();
+    }
+    await new Promise((resolve) => setTimeout(resolve, 0));
+    process.off('unhandledRejection', onUnhandled);
+    expect(unhandled).toHaveLength(0);
+  });
 
-    expect(agentService.getSession).toHaveBeenCalledWith(CONVERSATION_ID, {
+  it('when the 120s timeout fires while prompt() is still pending, returns timeout with one teardown, and a later prompt rejection surfaces no unhandledRejection (Task 5 finding 6 — timeout/prompt race)', async () => {
-      userId: USER_B.id,
+    const session = makeAgentSession(USER_A);
-      tenantId: USER_B.tenantId,
+    const detach = vi.fn();
+    let rejectPrompt: (reason: unknown) => void = () => undefined;
+    const prompting = new Promise<void>((_resolve, reject) => {
+      rejectPrompt = reject;
     });
-    expect(socket.emit).toHaveBeenCalledWith(
+    const svc = {
-      'error',
+      getSession: vi.fn(() => session),
-      expect.objectContaining({ conversationId: CONVERSATION_ID }),
+      createSession: vi.fn(),
-    );
+      onEvent: vi.fn(() => detach),
+      addChannel: vi.fn(),
+      removeChannel: vi.fn(),
+      prompt: vi.fn(() => prompting),
+      recordTokenUsage: vi.fn(),
+    };
+    const runtime = new EmbeddedChatRuntime(svc as never);
+
+    const unhandled: unknown[] = [];
+    const onUnhandled = (reason: unknown): void => {
+      unhandled.push(reason);
+    };
+    process.on('unhandledRejection', onUnhandled);
+    vi.useFakeTimers();
+    try {
+      const resultPromise = runtime.completeLegacyRestTurn(ctx, {
+        content: 'prompt settles after the deadline',
+      });
+      // The timeout wins the race while prompt is still pending.
+      await vi.advanceTimersByTimeAsync(200_000);
+      const result = await resultPromise;
+
+      expect(result).toEqual({ ok: false, code: 'timeout', retryable: true });
+      expect(detach).toHaveBeenCalledTimes(1);
+
+      // The prompt now rejects LATE — after the turn already returned its timeout result. Because
+      // Promise.all installed a rejection handler on `prompting` synchronously (the fix), this late
+      // rejection is already observed and must not escape as an unhandledRejection.
+      rejectPrompt(new Error('late backend failure'));
+    } finally {
+      vi.useRealTimers();
+    }
+    await new Promise((resolve) => setTimeout(resolve, 0));
+    process.off('unhandledRejection', onUnhandled);
+    expect(unhandled).toHaveLength(0);
   });
 });

apps/gateway/src/agent/routing/routing-e2e.test.ts

@@ -8,6 +8,7 @@
  * to avoid real I/O — they verify the complete classify → match → decide path.
  */
 import { describe, it, expect, vi } from 'vitest';
+import type { ProviderHealthStatus } from '@mosaicstack/types';
 import { RoutingEngineService } from './routing-engine.service.js';
 import { DEFAULT_ROUTING_RULES } from '../routing/default-rules.js';
 import type { RoutingRule } from './routing.types.js';
@@ -17,7 +18,7 @@ import type { RoutingRule } from './routing.types.js';
 /** Build a RoutingEngineService backed by the given rule set and health map. */
 function makeService(
   rules: RoutingRule[],
-  healthMap: Record<string, { status: string }>,
+  healthMap: Record<string, { status: ProviderHealthStatus }>,
 ): RoutingEngineService {
   const mockDb = {
     select: vi.fn().mockReturnValue({
@@ -67,11 +68,11 @@ function defaultRules(): RoutingRule[] {
 }
 
 /** A health map where anthropic, openai, and zai are all healthy. */
-const allHealthy: Record<string, { status: string }> = {
+const allHealthy: Record<string, { status: ProviderHealthStatus }> = {
-  anthropic: { status: 'up' },
+  anthropic: { status: 'healthy' },
-  openai: { status: 'up' },
+  openai: { status: 'healthy' },
-  zai: { status: 'up' },
+  zai: { status: 'healthy' },
-  ollama: { status: 'up' },
+  ollama: { status: 'healthy' },
 };
 
 // ─── M4-013 E2E tests ─────────────────────────────────────────────────────────
@@ -212,10 +213,10 @@ describe('M4-013: routing end-to-end pipeline', () => {
     // Let's use a simple coding message to target Simple coding → Codex (openai)
     const message = 'implement a sort function';
 
-    const unhealthyHealth = {
+    const unhealthyHealth: Record<string, { status: ProviderHealthStatus }> = {
       anthropic: { status: 'down' },
-      openai: { status: 'up' },
+      openai: { status: 'healthy' },
-      zai: { status: 'up' },
+      zai: { status: 'healthy' },
       ollama: { status: 'down' },
     };
 

apps/gateway/src/agent/routing/routing-engine.service.ts

@@ -1,5 +1,6 @@
 import { Inject, Injectable, Logger } from '@nestjs/common';
 import { routingRules, type Db, and, asc, eq, or } from '@mosaicstack/db';
+import type { ProviderHealthStatus } from '@mosaicstack/types';
 import { DB } from '../../database/database.module.js';
 import { ProviderService } from '../provider.service.js';
 import { classifyTask } from './task-classifier.js';
@@ -49,7 +50,7 @@ export class RoutingEngineService {
   async resolve(
     message: string,
     userId?: string,
-    availableProviders?: Record<string, { status: string }>,
+    availableProviders?: Record<string, { status: ProviderHealthStatus }>,
   ): Promise<RoutingDecision> {
     const classification = classifyTask(message);
     this.logger.debug(
@@ -69,9 +70,8 @@ export class RoutingEngineService {
       if (!this.matchConditions(rule, classification)) continue;
 
       const providerStatus = health[rule.action.provider]?.status;
-      const isHealthy = providerStatus === 'up' || providerStatus === 'ok';
 
-      if (!isHealthy) {
+      if (!this.isRoutable(providerStatus)) {
         this.logger.debug(
           `Rule "${rule.name}" matched but provider "${rule.action.provider}" is unhealthy (status: ${providerStatus ?? 'unknown'})`,
         );
@@ -111,6 +111,10 @@ export class RoutingEngineService {
 
   // ─── Private helpers ───────────────────────────────────────────────────────
 
+  private isRoutable(status: ProviderHealthStatus | undefined): boolean {
+    return status === 'healthy' || status === 'degraded';
+  }
+
   private evaluateCondition(
     condition: RoutingCondition,
     classification: TaskClassification,
@@ -186,11 +190,12 @@ export class RoutingEngineService {
    * Walk the fallback chain and return the first healthy provider/model pair.
    * If none are healthy, return the first entry unconditionally (last resort).
    */
-  private applyFallbackChain(health: Record<string, { status: string }>): RoutingDecision {
+  private applyFallbackChain(
+    health: Record<string, { status: ProviderHealthStatus }>,
+  ): RoutingDecision {
     for (const candidate of FALLBACK_CHAIN) {
       const providerStatus = health[candidate.provider]?.status;
-      const isHealthy = providerStatus === 'up' || providerStatus === 'ok';
+      if (this.isRoutable(providerStatus)) {
-      if (isHealthy) {
         this.logger.debug(`Fallback resolved: ${candidate.provider}/${candidate.model}`);
         return {
           provider: candidate.provider,

apps/gateway/src/agent/routing/routing-engine.test.ts

@@ -1,4 +1,5 @@
 import { describe, it, expect, vi, beforeEach } from 'vitest';
+import type { ProviderHealthStatus } from '@mosaicstack/types';
 import { RoutingEngineService } from './routing-engine.service.js';
 import type { RoutingRule, TaskClassification } from './routing.types.js';
 
@@ -29,7 +30,7 @@ function makeClassification(overrides: Partial<TaskClassification> = {}): TaskCl
 /** Build a minimal RoutingEngineService with mocked DB and ProviderService. */
 function makeService(
   rules: RoutingRule[] = [],
-  healthMap: Record<string, { status: string }> = {},
+  healthMap: Record<string, { status: ProviderHealthStatus }> = {},
 ): RoutingEngineService {
   const mockDb = {
     select: vi.fn().mockReturnValue({
@@ -217,7 +218,10 @@ describe('RoutingEngineService.resolve — priority ordering', () => {
       }),
     ];
 
-    const service = makeService(rules, { anthropic: { status: 'up' }, openai: { status: 'up' } });
+    const service = makeService(rules, {
+      anthropic: { status: 'healthy' },
+      openai: { status: 'healthy' },
+    });
 
     const decision = await service.resolve('implement a function');
     expect(decision.ruleName).toBe('high priority');
@@ -241,7 +245,10 @@ describe('RoutingEngineService.resolve — priority ordering', () => {
       }),
     ];
 
-    const service = makeService(rules, { anthropic: { status: 'up' }, openai: { status: 'up' } });
+    const service = makeService(rules, {
+      anthropic: { status: 'healthy' },
+      openai: { status: 'healthy' },
+    });
 
     const decision = await service.resolve('implement a function');
     expect(decision.ruleName).toBe('coding rule');
@@ -270,7 +277,7 @@ describe('RoutingEngineService.resolve — unhealthy provider handling', () => {
 
     const service = makeService(rules, {
       anthropic: { status: 'down' }, // primary is unhealthy
-      openai: { status: 'up' },
+      openai: { status: 'healthy' },
     });
 
     const decision = await service.resolve('implement a function');
@@ -290,7 +297,7 @@ describe('RoutingEngineService.resolve — unhealthy provider handling', () => {
     ];
 
     const service2 = makeService(unhealthyRules, {
-      anthropic: { status: 'up' },
+      anthropic: { status: 'healthy' },
       openai: { status: 'down' },
     });
 
@@ -306,7 +313,7 @@ describe('RoutingEngineService.resolve — unhealthy provider handling', () => {
 
     const service = makeService(rules, {
       anthropic: { status: 'down' }, // Sonnet is on anthropic — down
-      ollama: { status: 'up' }, // Haiku is also on anthropic — use Ollama as next
+      ollama: { status: 'healthy' }, // Haiku is also on anthropic — use Ollama as next
     });
 
     const decision = await service.resolve('hello there');
@@ -345,7 +352,7 @@ describe('RoutingEngineService.resolve — empty conditions (fallback rule)', ()
       }),
     ];
 
-    const service = makeService(rules, { anthropic: { status: 'up' } });
+    const service = makeService(rules, { anthropic: { status: 'healthy' } });
 
     const decision = await service.resolve('completely unrelated message xyz');
     expect(decision.ruleName).toBe('catch-all');
@@ -369,7 +376,7 @@ describe('RoutingEngineService.resolve — empty conditions (fallback rule)', ()
       }),
     ];
 
-    const service = makeService(rules, { anthropic: { status: 'up' } });
+    const service = makeService(rules, { anthropic: { status: 'healthy' } });
 
     const codingDecision = await service.resolve('implement a function');
     expect(codingDecision.ruleName).toBe('specific coding rule');
@@ -401,7 +408,7 @@ describe('RoutingEngineService.resolve — disabled rules', () => {
       }),
     ];
 
-    const service = makeService(rules, { anthropic: { status: 'up' } });
+    const service = makeService(rules, { anthropic: { status: 'healthy' } });
 
     const decision = await service.resolve('implement a function');
     expect(decision.ruleName).toBe('enabled fallback');
@@ -452,9 +459,45 @@ describe('RoutingEngineService.resolve — availableProviders override', () => {
       ps: unknown,
     ) => RoutingEngineService)(mockDb, mockProviderService);
 
-    const preSupplied = { anthropic: { status: 'up' } };
+    const preSupplied: Record<string, { status: ProviderHealthStatus }> = {
+      anthropic: { status: 'healthy' },
+    };
     await service.resolve('implement a function', undefined, preSupplied);
 
     expect(mockHealthCheckAll).not.toHaveBeenCalled();
   });
 });
+
+// ─── resolve — canonical ProviderHealthStatus values ──────────────────────────
+
+describe('RoutingEngineService.resolve — canonical health status routing', () => {
+  it('routes healthy and degraded providers by rule, and falls through to fallback when down', async () => {
+    const codingRule = makeRule({
+      name: 'coding rule',
+      priority: 1,
+      conditions: [{ field: 'taskType', operator: 'eq', value: 'coding' }],
+      action: { provider: 'openai', model: 'gpt-4o' },
+    });
+
+    // healthy → selected by its own rule, not the fallback chain
+    const healthyService = makeService([codingRule], { openai: { status: 'healthy' } });
+    const healthyDecision = await healthyService.resolve('implement a function');
+    expect(healthyDecision.ruleName).toBe('coding rule');
+    expect(healthyDecision.provider).toBe('openai');
+
+    // down → rule is skipped as unroutable, falls through to the fallback chain
+    const downService = makeService([codingRule], {
+      openai: { status: 'down' },
+      anthropic: { status: 'healthy' },
+    });
+    const downDecision = await downService.resolve('implement a function');
+    expect(downDecision.ruleName).toBe('fallback');
+    expect(downDecision.provider).toBe('anthropic');
+
+    // degraded → still routable, selected by its own rule, not the fallback chain
+    const degradedService = makeService([codingRule], { openai: { status: 'degraded' } });
+    const degradedDecision = await degradedService.resolve('implement a function');
+    expect(degradedDecision.ruleName).toBe('coding rule');
+    expect(degradedDecision.provider).toBe('openai');
+  });
+});

apps/gateway/src/app.module.spec.ts

@@ -0,0 +1,624 @@
+import 'reflect-metadata';
+import { mkdtemp, mkdir, readFile, rm, writeFile } from 'node:fs/promises';
+import * as nodeOs from 'node:os';
+import { dirname, isAbsolute, join, relative, resolve } from 'node:path';
+import * as nodeUrl from 'node:url';
+import { MODULE_METADATA } from '@nestjs/common/constants.js';
+import { describe, expect, it, vi } from 'vitest';
+import type { MosaicConfig } from '@mosaicstack/config';
+
+interface ComposedModuleGraph {
+  imports: readonly unknown[];
+  federationModule: unknown;
+  bootLogLines: readonly string[];
+  mosaicConfig: MosaicConfig;
+  resolvedConfigPath: string;
+}
+
+type StorageTier = 'local' | 'standalone' | 'federated';
+
+interface ModuleGraphFixture {
+  tempRoot: string;
+  anchor: string;
+  homePath: string;
+  cwdPath: string;
+  monorepoRootEnvPath: string;
+  gatewayLocalEnvPath: string;
+  daemonEnvPath: string;
+  monorepoRootConfigPath: string;
+  gatewayLocalConfigPath: string;
+}
+
+interface ModuleGraphFixtureOptions {
+  rootEnvMode?: 'present' | 'absent';
+  rootTier?: StorageTier;
+  rootEnvContents?: string;
+  redactionMarker?: string;
+  gatewayLocalTier?: StorageTier;
+  gatewayLocalEnvContents?: string;
+  daemonEnvContents?: string;
+  inheritedTier?: StorageTier;
+  expectedProcessTier?: string;
+  setup?: (fixture: ModuleGraphFixture) => Promise<void>;
+}
+
+// Each case uses vi.resetModules() and re-imports the full gateway graph for distinct ambient FS/env; CI needs headroom, while this still guards genuine hangs.
+const MODULE_IMPORT_TIMEOUT_MS = 120_000;
+const MONOREPO_ROOT_DOTENV_LABEL = 'monorepo-root .env';
+const DAEMON_DOTENV_LABEL = 'daemon .env';
+
+function configJson(tier: StorageTier): string {
+  if (tier === 'local') {
+    return JSON.stringify({
+      tier,
+      storage: { type: 'pglite', dataDir: '.mosaic/storage-pglite' },
+      queue: { type: 'local', dataDir: '.mosaic/queue' },
+      memory: { type: 'keyword' },
+    });
+  }
+
+  return JSON.stringify({
+    tier,
+    storage: { type: 'postgres', url: 'postgresql://fixture.invalid/mosaic' },
+    queue: { type: 'bullmq' },
+    memory: { type: tier === 'federated' ? 'pgvector' : 'keyword' },
+  });
+}
+
+function snapshotProcessEnv(): Record<string, string | undefined> {
+  return { ...process.env };
+}
+
+function restoreProcessEnv(snapshot: Record<string, string | undefined>): void {
+  for (const key of Object.keys(process.env)) {
+    if (!(key in snapshot)) {
+      delete process.env[key];
+    }
+  }
+
+  for (const [key, value] of Object.entries(snapshot)) {
+    if (value === undefined) {
+      delete process.env[key];
+      continue;
+    }
+
+    process.env[key] = value;
+  }
+}
+
+function expectPathUnderTempRoot(path: string, tempRoot: string): void {
+  const relativePath = relative(tempRoot, path);
+  expect(relativePath === '' || (!relativePath.startsWith('..') && !isAbsolute(relativePath))).toBe(
+    true,
+  );
+}
+
+async function writeFixture(path: string, contents: string, tempRoot: string): Promise<void> {
+  expectPathUnderTempRoot(path, tempRoot);
+  await mkdir(dirname(path), { recursive: true });
+  await writeFile(path, contents, 'utf8');
+}
+
+interface ConfigModuleProvider {
+  provide: string;
+  useFactory: () => MosaicConfig;
+}
+
+function isConfigModuleProvider(value: unknown): value is ConfigModuleProvider {
+  if (typeof value !== 'object' || value === null) {
+    return false;
+  }
+
+  if (!('provide' in value) || typeof value.provide !== 'string') {
+    return false;
+  }
+
+  return 'useFactory' in value && typeof value.useFactory === 'function';
+}
+
+function singleBootLogLine(bootLogLines: readonly string[]): string {
+  expect(bootLogLines).toHaveLength(1);
+  const [bootLogLine] = bootLogLines;
+  if (bootLogLine === undefined) {
+    throw new Error('Expected a single boot log line');
+  }
+
+  return bootLogLine;
+}
+
+function expectBootLogLine(
+  bootLogLines: readonly string[],
+  tier: StorageTier,
+  source: string,
+): void {
+  const bootLogLine = singleBootLogLine(bootLogLines);
+
+  expect(bootLogLine).toContain(`storage tier=${tier}`);
+  expect(bootLogLine).toContain(`source=${source}`);
+}
+
+async function loadModuleGraphFromDotenv(
+  options: ModuleGraphFixtureOptions,
+): Promise<ComposedModuleGraph> {
+  const originalEnv = snapshotProcessEnv();
+  const tempRoot = await mkdtemp(join(nodeOs.tmpdir(), 'mosaic-gateway-module-'));
+  let consoleInfoSpy: ReturnType<typeof vi.spyOn> | undefined;
+  let cwdSpy: ReturnType<typeof vi.spyOn> | undefined;
+
+  try {
+    const anchor = join(tempRoot, 'anchored', 'apps', 'gateway', 'src');
+    const homePath = join(tempRoot, 'home');
+    const cwdPath = join(tempRoot, 'ambient', 'parent', 'cwd');
+    const fixture: ModuleGraphFixture = {
+      tempRoot,
+      anchor,
+      homePath,
+      cwdPath,
+      monorepoRootEnvPath: resolve(anchor, '../../..', '.env'),
+      gatewayLocalEnvPath: resolve(anchor, '..', '.env'),
+      daemonEnvPath: join(homePath, '.config', 'mosaic', 'gateway', '.env'),
+      monorepoRootConfigPath: resolve(anchor, '../../..', 'mosaic.config.json'),
+      gatewayLocalConfigPath: resolve(anchor, '..', 'mosaic.config.json'),
+    };
+    consoleInfoSpy = vi.spyOn(console, 'info').mockImplementation((): void => undefined);
+
+    for (const path of Object.values(fixture)) {
+      expectPathUnderTempRoot(path, tempRoot);
+    }
+
+    await mkdir(anchor, { recursive: true });
+    await mkdir(cwdPath, { recursive: true });
+
+    if ((options.rootEnvMode ?? 'present') === 'absent') {
+      if (
+        options.rootEnvContents !== undefined ||
+        options.rootTier !== undefined ||
+        options.redactionMarker !== undefined
+      ) {
+        throw new Error('Expected no root env fixture values when rootEnvMode is absent');
+      }
+    } else {
+      if (options.rootEnvContents === undefined && options.rootTier === undefined) {
+        throw new Error('Expected rootTier or rootEnvContents');
+      }
+
+      const rootFixture = options.rootEnvContents ?? `MOSAIC_STORAGE_TIER=${options.rootTier}\n`;
+      const rootFixtureWithMarker = options.redactionMarker
+        ? `${rootFixture}BETTER_AUTH_SECRET=${options.redactionMarker}\n`
+        : rootFixture;
+      await writeFixture(fixture.monorepoRootEnvPath, rootFixtureWithMarker, tempRoot);
+    }
+
+    if (options.daemonEnvContents !== undefined) {
+      await writeFixture(fixture.daemonEnvPath, options.daemonEnvContents, tempRoot);
+    }
+
+    if (options.gatewayLocalEnvContents !== undefined) {
+      await writeFixture(fixture.gatewayLocalEnvPath, options.gatewayLocalEnvContents, tempRoot);
+    } else if (options.gatewayLocalTier !== undefined) {
+      await writeFixture(
+        fixture.gatewayLocalEnvPath,
+        `MOSAIC_STORAGE_TIER=${options.gatewayLocalTier}\n`,
+        tempRoot,
+      );
+    }
+
+    process.env['HOME'] = homePath;
+    delete process.env['MOSAIC_STORAGE_TIER'];
+    delete process.env['DATABASE_URL'];
+    delete process.env['VALKEY_URL'];
+    delete process.env['MOSAIC_GATEWAY_HOME'];
+
+    await options.setup?.(fixture);
+
+    if (options.inheritedTier !== undefined) {
+      process.env['MOSAIC_STORAGE_TIER'] = options.inheritedTier;
+    }
+
+    vi.resetModules();
+    vi.doMock('node:os', () => ({ ...nodeOs, homedir: (): string => homePath }));
+    vi.doMock('node:url', () => ({
+      ...nodeUrl,
+      fileURLToPath: (url: string | URL): string => {
+        const actualPath = nodeUrl.fileURLToPath(url);
+        if (
+          actualPath.endsWith('/apps/gateway/src/env.ts') ||
+          actualPath.endsWith('/apps/gateway/src/env.js')
+        ) {
+          return join(anchor, 'env.ts');
+        }
+        return actualPath;
+      },
+    }));
+    cwdSpy = vi.spyOn(process, 'cwd').mockReturnValue(cwdPath);
+
+    if (options.inheritedTier === undefined) {
+      expect(process.env['MOSAIC_STORAGE_TIER']).toBeUndefined();
+    } else {
+      expect(process.env['MOSAIC_STORAGE_TIER']).toBe(options.inheritedTier);
+    }
+
+    const envModule = await import('./env.js');
+    expect(process.env['MOSAIC_STORAGE_TIER']).toBe(
+      options.expectedProcessTier ?? options.rootTier,
+    );
+
+    const { AppModule } = await import('./app.module.js');
+    const { FederationModule } = await import('./federation/federation.module.js');
+    const imports: unknown = Reflect.getMetadata(MODULE_METADATA.IMPORTS, AppModule);
+
+    if (!Array.isArray(imports)) {
+      throw new Error('AppModule imports metadata is not an array');
+    }
+
+    const { ConfigModule, MOSAIC_CONFIG } = await import('./config/config.module.js');
+    const providers: unknown = Reflect.getMetadata(MODULE_METADATA.PROVIDERS, ConfigModule);
+
+    if (!Array.isArray(providers)) {
+      throw new Error('ConfigModule providers metadata is not an array');
+    }
+
+    const configProvider = providers
+      .filter(isConfigModuleProvider)
+      .find((provider: ConfigModuleProvider): boolean => provider.provide === MOSAIC_CONFIG);
+
+    if (!configProvider) {
+      throw new Error('MOSAIC_CONFIG provider factory not found');
+    }
+
+    return {
+      imports,
+      federationModule: FederationModule,
+      bootLogLines: consoleInfoSpy.mock.calls.map((args: readonly unknown[]): string =>
+        args.map((value: unknown): string => String(value)).join(' '),
+      ),
+      mosaicConfig: configProvider.useFactory(),
+      resolvedConfigPath: envModule.resolveGatewayConfigPath(),
+    };
+  } finally {
+    cwdSpy?.mockRestore();
+    vi.doUnmock('node:url');
+    vi.doUnmock('node:os');
+    vi.resetModules();
+    consoleInfoSpy?.mockRestore();
+    restoreProcessEnv(originalEnv);
+    await rm(tempRoot, { recursive: true, force: true });
+  }
+}
+
+describe('AppModule federation gating', (): void => {
+  it('loads dotenv before tracing and AppModule evaluation', async (): Promise<void> => {
+    const mainSource = await readFile(new URL('./main.ts', import.meta.url), 'utf8');
+    const envImportIndex = mainSource.indexOf("import './env.js';");
+    const tracingImportIndex = mainSource.indexOf("import './tracing.js';");
+    const appModuleImportIndex = mainSource.indexOf("import { AppModule } from './app.module.js';");
+
+    expect(envImportIndex).toBeGreaterThan(-1);
+    expect(envImportIndex).toBeLessThan(tracingImportIndex);
+    expect(envImportIndex).toBeLessThan(appModuleImportIndex);
+  });
+
+  it(
+    'ignores ambient cwd/.env and cwd/../.env files',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'local',
+        setup: async (fixture: ModuleGraphFixture): Promise<void> => {
+          await writeFixture(
+            join(fixture.cwdPath, '.env'),
+            'MOSAIC_STORAGE_TIER=federated\n',
+            fixture.tempRoot,
+          );
+          await writeFixture(
+            resolve(fixture.cwdPath, '..', '.env'),
+            'MOSAIC_STORAGE_TIER=federated\n',
+            fixture.tempRoot,
+          );
+        },
+      });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'local', MONOREPO_ROOT_DOTENV_LABEL);
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'ignores an ambient cwd/mosaic.config.json federated config',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'local',
+        setup: async (fixture: ModuleGraphFixture): Promise<void> => {
+          await writeFixture(
+            join(fixture.cwdPath, 'mosaic.config.json'),
+            configJson('federated'),
+            fixture.tempRoot,
+          );
+        },
+      });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'local', MONOREPO_ROOT_DOTENV_LABEL);
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'ignores an ambient cwd/../../mosaic.config.json federated config',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'local',
+        setup: async (fixture: ModuleGraphFixture): Promise<void> => {
+          await writeFixture(
+            resolve(fixture.cwdPath, '../..', 'mosaic.config.json'),
+            configJson('federated'),
+            fixture.tempRoot,
+          );
+        },
+      });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'local', MONOREPO_ROOT_DOTENV_LABEL);
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'anchored gateway-local config wins monorepo-root config and registers FederationModule',
+    async (): Promise<void> => {
+      let gatewayLocalConfigPath = '';
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'local',
+        setup: async (fixture: ModuleGraphFixture): Promise<void> => {
+          gatewayLocalConfigPath = fixture.gatewayLocalConfigPath;
+          await writeFixture(
+            fixture.gatewayLocalConfigPath,
+            configJson('federated'),
+            fixture.tempRoot,
+          );
+          await writeFixture(fixture.monorepoRootConfigPath, configJson('local'), fixture.tempRoot);
+        },
+      });
+
+      expect(graph.resolvedConfigPath).toBe(gatewayLocalConfigPath);
+      expect(graph.mosaicConfig.tier).toBe('federated');
+      expect(graph.imports).toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'federated', 'mosaic.config.json');
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'resolves the daemon-installed GATEWAY_HOME/mosaic.config.json ahead of gateway-local and monorepo-root configs',
+    async (): Promise<void> => {
+      let daemonConfigPath = '';
+      const graph = await loadModuleGraphFromDotenv({
+        rootEnvMode: 'absent',
+        setup: async (fixture: ModuleGraphFixture): Promise<void> => {
+          const externalGatewayHome = join(fixture.tempRoot, 'external-gateway-home');
+          daemonConfigPath = join(externalGatewayHome, 'mosaic.config.json');
+          await writeFixture(daemonConfigPath, configJson('federated'), fixture.tempRoot);
+          await writeFixture(
+            fixture.gatewayLocalConfigPath,
+            configJson('standalone'),
+            fixture.tempRoot,
+          );
+          await writeFixture(fixture.monorepoRootConfigPath, configJson('local'), fixture.tempRoot);
+          process.env['MOSAIC_GATEWAY_HOME'] = externalGatewayHome;
+          process.env['DATABASE_URL'] = 'postgresql://fixture.invalid/mosaic';
+        },
+      });
+
+      expect(graph.resolvedConfigPath).toBe(daemonConfigPath);
+      expect(graph.mosaicConfig.tier).toBe('federated');
+      expect(graph.imports).toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'federated', 'mosaic.config.json');
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'logs mosaic.config.json when anchored config and env tiers are both federated',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'federated',
+        setup: async (fixture: ModuleGraphFixture): Promise<void> => {
+          await writeFixture(
+            fixture.monorepoRootConfigPath,
+            configJson('federated'),
+            fixture.tempRoot,
+          );
+        },
+      });
+
+      expect(graph.imports).toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'federated', 'mosaic.config.json');
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'logs standalone from a monorepo-root .env DATABASE_URL fallback',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootEnvContents: 'DATABASE_URL=fixture-database-url\n',
+      });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'standalone', MONOREPO_ROOT_DOTENV_LABEL);
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'attributes an invalid monorepo-root dotenv tier to the default',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootEnvContents: 'MOSAIC_STORAGE_TIER=invalid\n',
+        expectedProcessTier: 'invalid',
+      });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'local', 'default');
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'attributes DATABASE_URL fallback to daemon .env ahead of inherited local tier',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootEnvMode: 'absent',
+        daemonEnvContents: 'DATABASE_URL=fixture-database-url\n',
+        inheritedTier: 'local',
+        expectedProcessTier: 'local',
+      });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'standalone', DAEMON_DOTENV_LABEL);
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'daemon .env wins over monorepo-root and gateway-local tier values',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'local',
+        gatewayLocalTier: 'federated',
+        daemonEnvContents: 'MOSAIC_STORAGE_TIER=standalone\n',
+        expectedProcessTier: 'standalone',
+      });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'standalone', DAEMON_DOTENV_LABEL);
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'inherits process.env.MOSAIC_STORAGE_TIER over daemon, monorepo-root, and gateway-local dotenv values',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'local',
+        gatewayLocalTier: 'federated',
+        daemonEnvContents: 'MOSAIC_STORAGE_TIER=federated\n',
+        inheritedTier: 'standalone',
+        expectedProcessTier: 'standalone',
+      });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'standalone', 'process environment');
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'gateway-local .env configures the tier and source when the monorepo-root .env is absent',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootEnvMode: 'absent',
+        gatewayLocalTier: 'federated',
+        expectedProcessTier: 'federated',
+      });
+
+      expect(graph.imports).toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'federated', 'gateway-local .env');
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'monorepo-root .env wins over gateway-local tier values',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'standalone',
+        gatewayLocalTier: 'federated',
+      });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'standalone', MONOREPO_ROOT_DOTENV_LABEL);
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it.each(['local', 'standalone'] as const)(
+    'does not register FederationModule for the %s tier',
+    async (tier): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({ rootTier: tier });
+
+      expect(graph.imports).not.toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, tier, MONOREPO_ROOT_DOTENV_LABEL);
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'registers FederationModule when federated tier is supplied by the anchored monorepo root .env',
+    async (): Promise<void> => {
+      const redactionMarker = 'redaction-fixture-marker';
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'federated',
+        redactionMarker,
+      });
+
+      expect(graph.imports).toContain(graph.federationModule);
+      expectBootLogLine(graph.bootLogLines, 'federated', MONOREPO_ROOT_DOTENV_LABEL);
+      expect(singleBootLogLine(graph.bootLogLines)).not.toContain(redactionMarker);
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'MOSAIC_CONFIG provider ignores an ambient cwd/mosaic.config.json config',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'local',
+        setup: async (fixture: ModuleGraphFixture): Promise<void> => {
+          await writeFixture(
+            join(fixture.cwdPath, 'mosaic.config.json'),
+            JSON.stringify({
+              tier: 'federated',
+              storage: {
+                type: 'postgres',
+                url: 'postgresql://ambient-attacker.invalid/mosaic',
+                enableVector: true,
+              },
+              queue: { type: 'bullmq' },
+              memory: { type: 'pgvector' },
+            }),
+            fixture.tempRoot,
+          );
+        },
+      });
+
+      expect(graph.mosaicConfig.tier).toBe('local');
+      expect(graph.mosaicConfig.storage).not.toEqual(
+        expect.objectContaining({ url: 'postgresql://ambient-attacker.invalid/mosaic' }),
+      );
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+
+  it(
+    'MOSAIC_CONFIG provider resolves from the anchored monorepo-root mosaic.config.json',
+    async (): Promise<void> => {
+      const graph = await loadModuleGraphFromDotenv({
+        rootTier: 'local',
+        setup: async (fixture: ModuleGraphFixture): Promise<void> => {
+          await writeFixture(
+            fixture.monorepoRootConfigPath,
+            configJson('federated'),
+            fixture.tempRoot,
+          );
+        },
+      });
+
+      expect(graph.mosaicConfig.tier).toBe('federated');
+      expect(graph.mosaicConfig.storage).toEqual(
+        expect.objectContaining({ url: 'postgresql://fixture.invalid/mosaic' }),
+      );
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+});

apps/gateway/src/app.module.ts

@@ -21,11 +21,22 @@ import { AdminModule } from './admin/admin.module.js';
 import { CommandsModule } from './commands/commands.module.js';
 import { PreferencesModule } from './preferences/preferences.module.js';
 import { GCModule } from './gc/gc.module.js';
+import { HarnessModule } from './harness/harness.module.js';
 import { ReloadModule } from './reload/reload.module.js';
 import { WorkspaceModule } from './workspace/workspace.module.js';
 import { QueueModule } from './queue/queue.module.js';
 import { FederationModule } from './federation/federation.module.js';
 import { ThrottlerGuard, ThrottlerModule } from '@nestjs/throttler';
+import { loadConfig } from '@mosaicstack/config';
+import { resolveGatewayConfigPath } from './env.js';
+
+// Federation (step-ca client, enrollment, federation verbs) is only wired for
+// tier 'federated' — CaService hard-requires STEP_CA_* at construction, which
+// must not gate standalone/local boots (docker-compose.federated.yml: the
+// federation profile "must not start in non-federated dev"). The gateway
+// entrypoint loads env.ts before evaluating this module so dotenv-backed tier
+// configuration is visible here.
+const federationEnabled = loadConfig(resolveGatewayConfigPath()).tier === 'federated';
 
 @Module({
   imports: [
@@ -50,10 +61,11 @@ import { ThrottlerGuard, ThrottlerModule } from '@nestjs/throttler';
     PreferencesModule,
     CommandsModule,
     GCModule,
+    HarnessModule,
     QueueModule,
     ReloadModule,
     WorkspaceModule,
-    FederationModule,
+    ...(federationEnabled ? [FederationModule] : []),
   ],
   controllers: [HealthController],
   providers: [

apps/gateway/src/chat/chat-runtime-router.spec.ts

@@ -0,0 +1,920 @@
+import 'reflect-metadata';
+import { Global, Module } from '@nestjs/common';
+import { Test, type TestingModule } from '@nestjs/testing';
+import { afterAll, afterEach, beforeAll, describe, expect, it } from 'vitest';
+import type { HarnessAdapter, HarnessConversationService } from '@mosaicstack/types';
+import { AgentService } from '../agent/agent.service.js';
+import { AuthGuard } from '../auth/auth.guard.js';
+import { CommandsModule } from '../commands/commands.module.js';
+import { HarnessModule } from '../harness/harness.module.js';
+import { ChatModule } from './chat.module.js';
+import { ChatGateway } from './chat.gateway.js';
+import { HarnessRegistry } from '../harness/harness.registry.js';
+import {
+  HARNESS_CONVERSATION_SERVICE,
+  HARNESS_CONVERSATION_SERVICE_UNAVAILABLE,
+  HARNESS_REGISTRY,
+  type HarnessConversationServiceBinding,
+} from '../harness/harness.tokens.js';
+import { ChatRuntimeRouter } from './chat-runtime-router.js';
+import {
+  ChatRuntimeUnavailableError,
+  ownConversation,
+  type ChatRuntime,
+  type ChatRuntimeMode,
+  type LegacyEmbeddedChatPort,
+  type LegacyRuntimeStream,
+  type LegacySessionPresentation,
+  type LegacySocketTurnLease,
+  type OwnedConversationContext,
+} from './chat-runtime.js';
+import { AppModule } from '../app.module.js';
+import { ProviderService } from '../agent/provider.service.js';
+
+/**
+ * Task Five, Step One (router). Proves the `ChatRuntimeRouter` resolves exactly one
+ * runtime by mode, fails closed at init when `pi-rpc` preconditions are unmet, and
+ * never downgrades `pi-rpc` to embedded execution. Red-first: the router is an
+ * unimplemented stub, so every behavioural assertion below fails until Step Three.
+ */
+
+const embedded: ChatRuntime = { kind: 'embedded' };
+const harness: ChatRuntime = { kind: 'harness' };
+
+/** A structurally-complete, non-sentinel conversation service. Its methods are never invoked here. */
+const boundConversationService = {
+  attach: () => Promise.reject(new Error('unused')),
+  detach: () => Promise.reject(new Error('unused')),
+  send: () => Promise.reject(new Error('unused')),
+
+  subscribeFrom: async function* () {
+    throw new Error('unused');
+  },
+} as unknown as HarnessConversationService;
+
+function registryWith(adapterIds: readonly string[]): HarnessRegistry {
+  const registry = new HarnessRegistry();
+  for (const id of adapterIds) {
+    registry.register({
+      id,
+      describe: () => Promise.reject(new Error('unused')),
+      catalog: () => Promise.reject(new Error('unused')),
+      create: () => Promise.reject(new Error('unused')),
+      resume: () => Promise.reject(new Error('unused')),
+    } as HarnessAdapter);
+  }
+  return registry;
+}
+
+function buildRouter(
+  mode: ChatRuntimeMode,
+  opts: { adapters: readonly string[]; service: HarnessConversationServiceBinding },
+): ChatRuntimeRouter {
+  return new ChatRuntimeRouter(registryWith(opts.adapters), opts.service, embedded, harness, mode);
+}
+
+/**
+ * Tear down a module that was deliberately driven to a fail-closed init.
+ * `NestApplicationContext.close()` re-awaits the module's `initializationPromise` before disposing
+ * (nest-application-context.js:127); when `init()` rejected, that await re-throws the SAME typed
+ * startup error, this time into teardown. Each caller here has already captured and asserted that
+ * exact `ChatRuntimeUnavailableError` via `initError`, so the re-throw is expected teardown noise —
+ * swallow ONLY that error, and surface anything else so a genuine teardown fault still fails loudly.
+ */
+async function closeIgnoringFailedInit(moduleRef: TestingModule): Promise<void> {
+  await moduleRef.close().catch((err: unknown) => {
+    if (err instanceof ChatRuntimeUnavailableError) return;
+    throw err;
+  });
+}
+
+describe('ChatRuntimeRouter', () => {
+  it('resolves only the harness runtime in pi-rpc mode when pi adapter and conversation service are present', () => {
+    const router = buildRouter('pi-rpc', {
+      adapters: ['pi'],
+      service: boundConversationService,
+    });
+
+    expect(() => router.onModuleInit()).not.toThrow();
+    expect(router.active).toBe(harness);
+    expect(router.active.kind).toBe('harness');
+  });
+
+  it('resolves only the embedded runtime in legacy mode and skips the pi preconditions', () => {
+    // Empty registry + unavailable service: legacy must ignore both and still start.
+    const router = buildRouter('legacy', {
+      adapters: [],
+      service: HARNESS_CONVERSATION_SERVICE_UNAVAILABLE,
+    });
+
+    expect(() => router.onModuleInit()).not.toThrow();
+    expect(router.active).toBe(embedded);
+    expect(router.active.kind).toBe('embedded');
+  });
+
+  it('fails closed at init when pi-rpc mode has no registered pi adapter', () => {
+    const router = buildRouter('pi-rpc', {
+      adapters: [],
+      service: boundConversationService,
+    });
+
+    expect(() => router.onModuleInit()).toThrow(ChatRuntimeUnavailableError);
+    try {
+      router.onModuleInit();
+      expect.unreachable('onModuleInit must throw when the pi adapter is absent');
+    } catch (err) {
+      expect(err).toBeInstanceOf(ChatRuntimeUnavailableError);
+      expect((err as ChatRuntimeUnavailableError).reason).toBe('adapter_unavailable');
+      expect((err as ChatRuntimeUnavailableError).code).toBe('runtime_unsupported');
+    }
+  });
+
+  it('fails closed at init when pi-rpc mode has the unavailable conversation-service sentinel', () => {
+    const router = buildRouter('pi-rpc', {
+      adapters: ['pi'],
+      service: HARNESS_CONVERSATION_SERVICE_UNAVAILABLE,
+    });
+
+    try {
+      router.onModuleInit();
+      expect.unreachable('onModuleInit must throw when the conversation service is unbound');
+    } catch (err) {
+      expect(err).toBeInstanceOf(ChatRuntimeUnavailableError);
+      expect((err as ChatRuntimeUnavailableError).reason).toBe('conversation_service_unavailable');
+      expect((err as ChatRuntimeUnavailableError).code).toBe('runtime_unsupported');
+    }
+  });
+
+  it('never falls back to embedded execution when pi-rpc preconditions are unmet', () => {
+    const router = buildRouter('pi-rpc', {
+      adapters: [],
+      service: HARNESS_CONVERSATION_SERVICE_UNAVAILABLE,
+    });
+
+    expect(() => router.onModuleInit()).toThrow(ChatRuntimeUnavailableError);
+    // A failed pi-rpc init must not silently expose the embedded runtime.
+    expect(() => router.active).toThrow();
+    let leaked: ChatRuntime | undefined;
+    try {
+      leaked = router.active;
+    } catch {
+      leaked = undefined;
+    }
+    expect(leaked).not.toBe(embedded);
+  });
+
+  it('exposes only fixed, browser-safe failure text (no raw provider or exception detail)', () => {
+    const router = buildRouter('pi-rpc', {
+      adapters: [],
+      service: boundConversationService,
+    });
+
+    try {
+      router.onModuleInit();
+      expect.unreachable('onModuleInit must throw');
+    } catch (err) {
+      const message = (err as ChatRuntimeUnavailableError).message;
+      expect(message).toBe(
+        'The pi-rpc chat runtime is unavailable: no "pi" harness adapter is registered.',
+      );
+      expect(message).not.toMatch(/Error:|\bat \b|node_modules|Symbol\(/);
+    }
+  });
+});
+
+/**
+ * Task Five, Step Three — legacy port operations fail closed under pi-rpc (direct valid-input).
+ *
+ * The unit suite above constructs the router but never invokes a legacy port operation, so the
+ * six per-operation inner `if (this.mode === 'pi-rpc')` guards are unexercised — a mutation that
+ * deletes one of them SURVIVES for lack of a test that drives that operation. This group closes
+ * that gap the right way: it drives each of the six operations DIRECTLY, in pi-rpc mode, with a
+ * valid branded {@link OwnedConversationContext} and valid input, against a recording embedded
+ * stub whose method returns a distinguishable `ok:true` success and increments a per-op counter.
+ *
+ * For each operation:
+ *  - pi-rpc test asserts the exact frozen `{ ok:false, code:'runtime_unsupported', retryable:false }`
+ *    result AND that the embedded stub was touched zero times (no effects);
+ *  - the paired legacy test proves that same stub method IS reached and returns its distinguishable
+ *    success when the mode does not refuse — so the pi-rpc zero-invocation assertion is meaningful,
+ *    not vacuously true because the stub could never be called.
+ *
+ * Deleting ONLY one operation's inner guard makes THAT operation's pi-rpc test behaviorally RED
+ * (the router returns the embedded `ok:true` value and records the call), with every outer guard
+ * and the other five inner guards intact. `next` is untouched; nothing here changes production.
+ */
+describe('ChatRuntimeRouter — legacy port ops fail closed under pi-rpc (Task Five, Step Three)', () => {
+  const RUNTIME_UNSUPPORTED = {
+    ok: false,
+    code: 'runtime_unsupported',
+    retryable: false,
+  } as const;
+
+  const PRESENTATION: LegacySessionPresentation = {
+    provider: 'embedded-provider',
+    modelId: 'embedded-model',
+    thinkingLevel: 'low',
+    availableThinkingLevels: ['low', 'high'],
+  };
+
+  const stream: LegacyRuntimeStream = {
+    channelId: 'websocket:test-socket',
+    onEvent: () => {},
+  };
+
+  const ctx = (): OwnedConversationContext =>
+    ownConversation('conversation-1', { userId: 'user-1', tenantId: 'tenant-1' });
+
+  /**
+   * Per-operation invocation counters with declared keys (not an index signature) so each
+   * `calls.<op>` is definitely `number` under `noUncheckedIndexedAccess`.
+   */
+  type LegacyPortCallCounts = {
+    completeLegacyRestTurn: number;
+    prepareLegacySocketTurn: number;
+    setLegacyThinking: number;
+    abortLegacyTurn: number;
+    applyLegacyModelOverride: number;
+    readLegacySessionPresentation: number;
+    dispatchVerifiedDiscordIngress: number;
+  };
+
+  /**
+   * An embedded port that records every invocation and returns a distinguishable `ok:true`
+   * value per operation. If a router op reaches it (its guard removed), both the recorded call
+   * count and the returned `ok:true` value diverge from the frozen `runtime_unsupported` result.
+   */
+  function recordingEmbeddedPort(): {
+    port: ChatRuntime & LegacyEmbeddedChatPort;
+    calls: LegacyPortCallCounts;
+  } {
+    const calls: LegacyPortCallCounts = {
+      completeLegacyRestTurn: 0,
+      prepareLegacySocketTurn: 0,
+      setLegacyThinking: 0,
+      abortLegacyTurn: 0,
+      applyLegacyModelOverride: 0,
+      readLegacySessionPresentation: 0,
+      dispatchVerifiedDiscordIngress: 0,
+    };
+    const lease: LegacySocketTurnLease = {
+      presentation: PRESENTATION,
+      dispatch: () => Promise.resolve({ ok: true, value: undefined }),
+      dispose: () => Promise.resolve(),
+    };
+    const port: ChatRuntime & LegacyEmbeddedChatPort = {
+      kind: 'embedded',
+      completeLegacyRestTurn: () => {
+        calls.completeLegacyRestTurn += 1;
+        return Promise.resolve({
+          ok: true,
+          value: { text: 'EMBEDDED-REST', presentation: PRESENTATION },
+        });
+      },
+      prepareLegacySocketTurn: () => {
+        calls.prepareLegacySocketTurn += 1;
+        return Promise.resolve({ ok: true, value: lease });
+      },
+      setLegacyThinking: () => {
+        calls.setLegacyThinking += 1;
+        return { ok: true, value: PRESENTATION };
+      },
+      abortLegacyTurn: () => {
+        calls.abortLegacyTurn += 1;
+        return Promise.resolve({ ok: true, value: undefined });
+      },
+      applyLegacyModelOverride: () => {
+        calls.applyLegacyModelOverride += 1;
+        return { ok: true, value: PRESENTATION };
+      },
+      readLegacySessionPresentation: () => {
+        calls.readLegacySessionPresentation += 1;
+        return { ok: true, value: PRESENTATION };
+      },
+      dispatchVerifiedDiscordIngress: () => {
+        calls.dispatchVerifiedDiscordIngress += 1;
+        return Promise.resolve({
+          ok: true,
+          value: {
+            presentation: PRESENTATION,
+            dispatch: () => Promise.resolve({ ok: true, value: undefined }),
+            dispose: () => Promise.resolve(),
+          },
+        });
+      },
+    };
+    return { port, calls };
+  }
+
+  function piRouter(port: ChatRuntime & LegacyEmbeddedChatPort): ChatRuntimeRouter {
+    return new ChatRuntimeRouter(
+      registryWith(['pi']),
+      boundConversationService,
+      port,
+      harness,
+      'pi-rpc',
+    );
+  }
+  function legacyRouter(port: ChatRuntime & LegacyEmbeddedChatPort): ChatRuntimeRouter {
+    return new ChatRuntimeRouter(
+      registryWith([]),
+      boundConversationService,
+      port,
+      harness,
+      'legacy',
+    );
+  }
+
+  // completeLegacyRestTurn ---------------------------------------------------
+  it('completeLegacyRestTurn refuses with runtime_unsupported and never touches embedded under pi-rpc', async () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = await piRouter(port).completeLegacyRestTurn(ctx(), { content: 'hello' });
+    expect(result).toEqual(RUNTIME_UNSUPPORTED);
+    expect(calls.completeLegacyRestTurn).toBe(0);
+  });
+  it('completeLegacyRestTurn delegates to embedded under legacy (guard is the sole gate)', async () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = await legacyRouter(port).completeLegacyRestTurn(ctx(), { content: 'hello' });
+    expect(result.ok).toBe(true);
+    expect(calls.completeLegacyRestTurn).toBe(1);
+  });
+
+  // prepareLegacySocketTurn --------------------------------------------------
+  it('prepareLegacySocketTurn refuses with runtime_unsupported and never touches embedded under pi-rpc', async () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = await piRouter(port).prepareLegacySocketTurn(
+      ctx(),
+      { content: 'hello' },
+      stream,
+    );
+    expect(result).toEqual(RUNTIME_UNSUPPORTED);
+    expect(calls.prepareLegacySocketTurn).toBe(0);
+  });
+  it('prepareLegacySocketTurn delegates to embedded under legacy (guard is the sole gate)', async () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = await legacyRouter(port).prepareLegacySocketTurn(
+      ctx(),
+      { content: 'hello' },
+      stream,
+    );
+    expect(result.ok).toBe(true);
+    expect(calls.prepareLegacySocketTurn).toBe(1);
+  });
+
+  // setLegacyThinking (sync) -------------------------------------------------
+  it('setLegacyThinking refuses with runtime_unsupported and never touches embedded under pi-rpc', () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = piRouter(port).setLegacyThinking(ctx(), 'high');
+    expect(result).toEqual(RUNTIME_UNSUPPORTED);
+    expect(calls.setLegacyThinking).toBe(0);
+  });
+  it('setLegacyThinking delegates to embedded under legacy (guard is the sole gate)', () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = legacyRouter(port).setLegacyThinking(ctx(), 'high');
+    expect(result.ok).toBe(true);
+    expect(calls.setLegacyThinking).toBe(1);
+  });
+
+  // abortLegacyTurn ----------------------------------------------------------
+  it('abortLegacyTurn refuses with runtime_unsupported and never touches embedded under pi-rpc', async () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = await piRouter(port).abortLegacyTurn(ctx());
+    expect(result).toEqual(RUNTIME_UNSUPPORTED);
+    expect(calls.abortLegacyTurn).toBe(0);
+  });
+  it('abortLegacyTurn delegates to embedded under legacy (guard is the sole gate)', async () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = await legacyRouter(port).abortLegacyTurn(ctx());
+    expect(result.ok).toBe(true);
+    expect(calls.abortLegacyTurn).toBe(1);
+  });
+
+  // applyLegacyModelOverride (sync) ------------------------------------------
+  it('applyLegacyModelOverride refuses with runtime_unsupported and never touches embedded under pi-rpc', () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = piRouter(port).applyLegacyModelOverride(ctx(), 'model-x');
+    expect(result).toEqual(RUNTIME_UNSUPPORTED);
+    expect(calls.applyLegacyModelOverride).toBe(0);
+  });
+  it('applyLegacyModelOverride delegates to embedded under legacy (guard is the sole gate)', () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = legacyRouter(port).applyLegacyModelOverride(ctx(), 'model-x');
+    expect(result.ok).toBe(true);
+    expect(calls.applyLegacyModelOverride).toBe(1);
+  });
+
+  // readLegacySessionPresentation (sync) -------------------------------------
+  it('readLegacySessionPresentation refuses with runtime_unsupported and never touches embedded under pi-rpc', () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = piRouter(port).readLegacySessionPresentation(ctx());
+    expect(result).toEqual(RUNTIME_UNSUPPORTED);
+    expect(calls.readLegacySessionPresentation).toBe(0);
+  });
+  it('readLegacySessionPresentation delegates to embedded under legacy (guard is the sole gate)', () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const result = legacyRouter(port).readLegacySessionPresentation(ctx());
+    expect(result.ok).toBe(true);
+    expect(calls.readLegacySessionPresentation).toBe(1);
+  });
+
+  // dispatchVerifiedDiscordIngress delegates in BOTH modes (embedded-only, no guard) ---------
+  it('dispatchVerifiedDiscordIngress delegates to embedded under pi-rpc (embedded-only, no mode guard)', async () => {
+    const { port, calls } = recordingEmbeddedPort();
+    const discordCtx = ctx() as unknown as Parameters<
+      ChatRuntimeRouter['dispatchVerifiedDiscordIngress']
+    >[0];
+    const result = await piRouter(port).dispatchVerifiedDiscordIngress(discordCtx, stream);
+    expect(result.ok).toBe(true);
+    expect(calls.dispatchVerifiedDiscordIngress).toBe(1);
+  });
+});
+
+/**
+ * Task Five, Step Two — group 1 (real Nest module-graph readiness).
+ *
+ * The unit suite above constructs the router directly. This group drives the SAME contract
+ * through a real NestJS graph: it imports the production `HarnessModule` (the proven-booting
+ * idiom from harness.controller.spec.ts) so the router resolves the REAL, empty `HarnessRegistry`
+ * via the real `HARNESS_REGISTRY` token, then runs the router's `OnModuleInit` through the Nest
+ * lifecycle (`moduleRef.init()`). Red-first: the router is an unimplemented stub whose
+ * `onModuleInit` throws a generic Error, so:
+ *   - readiness cases fail because the graph never comes up (init rejects), and
+ *   - fail-closed cases fail because a generic stub throw is NOT the SPECIFIC typed
+ *     `ChatRuntimeUnavailableError` (reason/code) the contract demands — a stub that
+ *     "throws anything" cannot mask these greens.
+ * The router is NOT wired into a production module yet, so it is provided here via a factory
+ * over the real registry token. Importing the real `ChatModule` bare is deliberately avoided:
+ * it injects `AgentService` without importing `AgentModule`, so its graph fails to RESOLVE — a
+ * collection/DI error, not a behavioural red. `next` is untouched; nothing here implements the router.
+ */
+describe('ChatRuntimeRouter — real Nest module-graph readiness (Task Five, Step Two group 1)', () => {
+  async function bootRouterGraph(
+    mode: ChatRuntimeMode,
+    opts: { adapters: readonly string[]; service: HarnessConversationServiceBinding },
+  ) {
+    const moduleRef = await Test.createTestingModule({
+      imports: [HarnessModule],
+      providers: [
+        {
+          provide: ChatRuntimeRouter,
+          useFactory: (registry: HarnessRegistry) =>
+            new ChatRuntimeRouter(registry, opts.service, embedded, harness, mode),
+          inject: [HARNESS_REGISTRY],
+        },
+      ],
+    })
+      // The imported HarnessModule's controllers reference AuthGuard (an HTTP-only concern,
+      // never exercised here); stub it so the graph resolves. The registry is NOT overridden —
+      // group 1 asserts against the genuine production HarnessRegistry.
+      .overrideGuard(AuthGuard)
+      .useValue({ canActivate: () => true })
+      .compile();
+
+    // Resolve the production registry singleton and register the requested adapters ON IT, so
+    // the router (which injects the same singleton) sees them when its lifecycle hook runs.
+    const registry = moduleRef.get<HarnessRegistry>(HARNESS_REGISTRY, { strict: false });
+    for (const id of opts.adapters) {
+      registry.register({
+        id,
+        describe: () => Promise.reject(new Error('unused')),
+        catalog: () => Promise.reject(new Error('unused')),
+        create: () => Promise.reject(new Error('unused')),
+        resume: () => Promise.reject(new Error('unused')),
+      } as HarnessAdapter);
+    }
+    return moduleRef;
+  }
+
+  // Capture an init rejection without letting a resolved init masquerade as success.
+  const initError = (moduleRef: { init(): Promise<unknown> }): Promise<unknown> =>
+    moduleRef.init().then(
+      () => new Error('module init resolved but the contract requires it to reject'),
+      (err: unknown) => err,
+    );
+
+  it('brings the graph up and resolves only the harness runtime in pi-rpc mode (pi adapter + bound service)', async () => {
+    const moduleRef = await bootRouterGraph('pi-rpc', {
+      adapters: ['pi'],
+      service: boundConversationService,
+    });
+    try {
+      await moduleRef.init();
+      const router = moduleRef.get(ChatRuntimeRouter, { strict: false });
+      expect(router.active).toBe(harness);
+      expect(router.active.kind).toBe('harness');
+    } finally {
+      await moduleRef.close();
+    }
+  });
+
+  it('brings the graph up in legacy mode over the REAL empty HarnessRegistry and resolves only the embedded runtime', async () => {
+    const moduleRef = await bootRouterGraph('legacy', {
+      adapters: [],
+      service: HARNESS_CONVERSATION_SERVICE_UNAVAILABLE,
+    });
+    try {
+      // Defense-in-depth: the production module wires the genuine registry, empty by default —
+      // guards against a test-double registry silently satisfying the readiness check.
+      const registry = moduleRef.get<HarnessRegistry>(HARNESS_REGISTRY, { strict: false });
+      expect(registry).toBeInstanceOf(HarnessRegistry);
+      expect(registry.list()).toHaveLength(0);
+
+      await moduleRef.init();
+      const router = moduleRef.get(ChatRuntimeRouter, { strict: false });
+      expect(router.active).toBe(embedded);
+      expect(router.active.kind).toBe('embedded');
+    } finally {
+      await moduleRef.close();
+    }
+  });
+
+  it('fails closed at module init when pi-rpc mode has no registered pi adapter (specific typed error, not a stub throw)', async () => {
+    const moduleRef = await bootRouterGraph('pi-rpc', {
+      adapters: [],
+      service: boundConversationService,
+    });
+    try {
+      const err = await initError(moduleRef);
+      expect(err).toBeInstanceOf(ChatRuntimeUnavailableError);
+      expect((err as ChatRuntimeUnavailableError).reason).toBe('adapter_unavailable');
+      expect((err as ChatRuntimeUnavailableError).code).toBe('runtime_unsupported');
+    } finally {
+      await closeIgnoringFailedInit(moduleRef);
+    }
+  });
+
+  it('fails closed at module init when pi-rpc mode has the unavailable conversation-service sentinel', async () => {
+    const moduleRef = await bootRouterGraph('pi-rpc', {
+      adapters: ['pi'],
+      service: HARNESS_CONVERSATION_SERVICE_UNAVAILABLE,
+    });
+    try {
+      const err = await initError(moduleRef);
+      expect(err).toBeInstanceOf(ChatRuntimeUnavailableError);
+      expect((err as ChatRuntimeUnavailableError).reason).toBe('conversation_service_unavailable');
+      expect((err as ChatRuntimeUnavailableError).code).toBe('runtime_unsupported');
+    } finally {
+      await closeIgnoringFailedInit(moduleRef);
+    }
+  });
+
+  it('surfaces only fixed, browser-safe failure text when the graph fails closed (no stub/exception detail)', async () => {
+    const moduleRef = await bootRouterGraph('pi-rpc', {
+      adapters: [],
+      service: boundConversationService,
+    });
+    try {
+      const err = await initError(moduleRef);
+      expect(err).toBeInstanceOf(ChatRuntimeUnavailableError);
+      const message = (err as ChatRuntimeUnavailableError).message;
+      expect(message).toBe(
+        'The pi-rpc chat runtime is unavailable: no "pi" harness adapter is registered.',
+      );
+      expect(message).not.toMatch(/Error:|\bat \b|node_modules|Symbol\(|not implemented/);
+    } finally {
+      await closeIgnoringFailedInit(moduleRef);
+    }
+  });
+});
+
+/**
+ * Task Five, Step Two — group 1b (production ChatModule wiring, declaration proof).
+ *
+ * Correction #1 (Scrappy fe3e02) asked for a red that imports the real `ChatModule` and calls
+ * `module.init()`. Investigated and found impractical/masking-prone: `ChatModule` provides
+ * `ChatGateway`, whose 10-argument constructor injects app-global providers (AgentService, AUTH,
+ * BRAIN, RoutingEngineService) plus the Commands/GC/Mcp/Reload subsystems across a forwardRef
+ * cycle. Booting it in isolation is a full-app integration boot — "override only unrelated
+ * dependencies" balloons into faking ~4 subsystems, and `overrideProvider` cannot even grant the
+ * cross-module export-scope visibility ChatGateway needs (probe: `ChatGateway` unresolved at
+ * `CommandExecutorService`). That is exactly the STOP-and-return branch of the directive.
+ *
+ * The faithful, unmaskable cover instead of a fragile boot: read the PRODUCTION `ChatModule`'s own
+ * Nest `@Module` metadata to prove it DECLARES the exclusive router provider and imports the real
+ * `HarnessModule` (the genuine registry source). This inspects the actual module object — not
+ * source text, not a test factory — so nothing can mask it. Group 1 above separately proves the
+ * router RESOLVES against the real, empty `HarnessRegistry` through the Nest lifecycle; the union
+ * of the two covers "the router is wired through ChatModule to the real registry" without the
+ * impractical single-graph boot. RED today (ChatModule provides only ChatGateway and imports only
+ * CommandsModule); GREEN once Step Three registers the router and imports HarnessModule.
+ */
+describe('ChatModule production wiring (Task Five, Step Two group 1b — declaration proof)', () => {
+  // Unwrap a forwardRef(() => Module) import to the module it references; pass others through.
+  const resolveImport = (imp: unknown): unknown =>
+    imp &&
+    typeof imp === 'object' &&
+    typeof (imp as { forwardRef?: unknown }).forwardRef === 'function'
+      ? (imp as { forwardRef: () => unknown }).forwardRef()
+      : imp;
+
+  // A provider entry is either a class (shorthand) or a { provide, ... } object; take its token.
+  const providerToken = (provider: unknown): unknown =>
+    typeof provider === 'function' ? provider : (provider as { provide?: unknown })?.provide;
+
+  it('declares the exclusive ChatRuntimeRouter as a provider on the production ChatModule', () => {
+    const providers: unknown[] = Reflect.getMetadata('providers', ChatModule) ?? [];
+    expect(providers.map(providerToken)).toContain(ChatRuntimeRouter);
+  });
+
+  it('imports the real HarnessModule into the production ChatModule (registry source, not a test double)', () => {
+    const imports: unknown[] = Reflect.getMetadata('imports', ChatModule) ?? [];
+    expect(imports.map(resolveImport)).toContain(HarnessModule);
+  });
+});
+
+/**
+ * Task Five, Step Two — group 1c (bounded real-`ChatModule` boot).
+ *
+ * Scrappy adjudication d67d2b (option c): boot the ACTUAL production `ChatModule` as the SUT and
+ * assert the exclusive router resolves THROUGH it — the single-graph proof group 1 (router over the
+ * real registry) and group 1b (production-module metadata) each cover only a half of. The heavy,
+ * UNRELATED cycle is the only thing bounded away, per the established isolation pattern in
+ * `apps/gateway/src/agent/hermes-runtime-reachability.e2e.test.ts`:
+ *   - `CommandsModule` (drags the Commands <-> Reload <-> Chat forwardRef cycle plus GC/Mcp/queue)
+ *     is replaced wholesale with an empty module via `.overrideModule(...).useModule(...)`;
+ *   - `ChatGateway` (10-arg constructor, an HTTP/socket concern never exercised here) is replaced
+ *     with an inert value;
+ *   - the sole legacy-controller dependency, `AgentService`, is supplied by a tiny `@Global()` stub;
+ *   - the HTTP-only `AuthGuard` is stubbed.
+ * Nothing about the router, `HarnessModule`, the registry, or the conversation-service binding is
+ * faked in the production-legacy case — those are retrieved from the REAL `ChatModule` graph. Mode
+ * is driven only through the production `CHAT_HARNESS_RUNTIME` env contract (`resolveChatRuntimeMode`).
+ *
+ * Red-first: today `ChatModule` neither imports `HarnessModule` nor provides `ChatRuntimeRouter`, so
+ * the booted graph contains no router/registry/conversation-service tokens. `init()` may resolve
+ * (there is no router lifecycle hook yet to reject), so every case fails on the MISSING actual
+ * router/registry/service wiring — not on unrelated DI, which is bounded away. GREEN at Step Three
+ * once `ChatModule` imports `HarnessModule`, provides the exclusive router, and binds the
+ * conversation-service token (defaulting to the unavailable sentinel).
+ */
+describe('ChatModule bounded real boot (Task Five, Step Two group 1c)', () => {
+  // The unrelated heavy cycle, replaced wholesale — not stubbed provider-by-provider.
+  @Module({})
+  class EmptyCommandsModule {}
+
+  // The ONLY genuine legacy dependency of the real ChatController, supplied inertly and globally so
+  // the pre-refactor controller instantiates without dragging AgentModule into the graph.
+  @Global()
+  @Module({
+    providers: [{ provide: AgentService, useValue: {} }],
+    exports: [AgentService],
+  })
+  class LegacyControllerDepsModule {}
+
+  const ORIGINAL_RUNTIME_ENV = process.env['CHAT_HARNESS_RUNTIME'];
+  afterEach(() => {
+    if (ORIGINAL_RUNTIME_ENV === undefined) delete process.env['CHAT_HARNESS_RUNTIME'];
+    else process.env['CHAT_HARNESS_RUNTIME'] = ORIGINAL_RUNTIME_ENV;
+  });
+
+  /**
+   * Boot the real ChatModule with only the unrelated cycle bounded away. `mode` is set through the
+   * genuine production env contract before providers instantiate. The optional overrides replace
+   * the registry / conversation-service the router injects, exercising the pi-rpc precondition
+   * branches through the ACTUAL module (they are no-ops today because those tokens are not yet in
+   * the graph — which is exactly why the router-retrieval assertions go red).
+   */
+  async function bootChatModule(
+    mode: ChatRuntimeMode,
+    overrides: {
+      registryAdapters?: readonly string[];
+      conversationService?: HarnessConversationServiceBinding;
+    } = {},
+  ): Promise<TestingModule> {
+    if (mode === 'pi-rpc') process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
+    else delete process.env['CHAT_HARNESS_RUNTIME'];
+
+    let builder = Test.createTestingModule({
+      imports: [LegacyControllerDepsModule, ChatModule],
+    })
+      .overrideModule(CommandsModule)
+      .useModule(EmptyCommandsModule)
+      .overrideProvider(ChatGateway)
+      .useValue({})
+      .overrideGuard(AuthGuard)
+      .useValue({ canActivate: () => true });
+
+    if (overrides.registryAdapters) {
+      builder = builder
+        .overrideProvider(HARNESS_REGISTRY)
+        .useValue(registryWith(overrides.registryAdapters));
+    }
+    if (overrides.conversationService !== undefined) {
+      builder = builder
+        .overrideProvider(HARNESS_CONVERSATION_SERVICE)
+        .useValue(overrides.conversationService);
+    }
+    return builder.compile();
+  }
+
+  // Capture an init rejection without letting a resolved init masquerade as success.
+  const initError = (moduleRef: TestingModule): Promise<unknown> =>
+    moduleRef.init().then(
+      () => new Error('module init resolved but the contract requires it to reject'),
+      (err: unknown) => err,
+    );
+
+  it('legacy mode: the actual router resolves the embedded runtime, the actual registry is empty, and the conversation-service token is the unavailable sentinel', async () => {
+    const moduleRef = await bootChatModule('legacy');
+    try {
+      await moduleRef.init();
+      const router = moduleRef.get(ChatRuntimeRouter, { strict: false });
+      expect(router.active.kind).toBe('embedded');
+
+      const registry = moduleRef.get<HarnessRegistry>(HARNESS_REGISTRY, { strict: false });
+      expect(registry).toBeInstanceOf(HarnessRegistry);
+      expect(registry.list()).toHaveLength(0);
+
+      const service = moduleRef.get<HarnessConversationServiceBinding>(
+        HARNESS_CONVERSATION_SERVICE,
+        {
+          strict: false,
+        },
+      );
+      expect(service).toBe(HARNESS_CONVERSATION_SERVICE_UNAVAILABLE);
+    } finally {
+      await moduleRef.close();
+    }
+  });
+
+  it('pi-rpc mode over the REAL empty registry fails closed at init with the typed adapter-unavailable error', async () => {
+    const moduleRef = await bootChatModule('pi-rpc');
+    try {
+      const err = await initError(moduleRef);
+      expect(err).toBeInstanceOf(ChatRuntimeUnavailableError);
+      expect((err as ChatRuntimeUnavailableError).reason).toBe('adapter_unavailable');
+      expect((err as ChatRuntimeUnavailableError).code).toBe('runtime_unsupported');
+    } finally {
+      await closeIgnoringFailedInit(moduleRef);
+    }
+  });
+
+  it('pi-rpc mode with a pi adapter present but the sentinel conversation service fails closed with the typed conversation-service-unavailable error', async () => {
+    const moduleRef = await bootChatModule('pi-rpc', {
+      registryAdapters: ['pi'],
+      conversationService: HARNESS_CONVERSATION_SERVICE_UNAVAILABLE,
+    });
+    try {
+      const err = await initError(moduleRef);
+      expect(err).toBeInstanceOf(ChatRuntimeUnavailableError);
+      expect((err as ChatRuntimeUnavailableError).reason).toBe('conversation_service_unavailable');
+      expect((err as ChatRuntimeUnavailableError).code).toBe('runtime_unsupported');
+    } finally {
+      await closeIgnoringFailedInit(moduleRef);
+    }
+  });
+
+  it('pi-rpc mode with a pi adapter and a bound conversation service: the actual router selects the harness runtime', async () => {
+    const moduleRef = await bootChatModule('pi-rpc', {
+      registryAdapters: ['pi'],
+      conversationService: boundConversationService,
+    });
+    try {
+      await moduleRef.init();
+      const router = moduleRef.get(ChatRuntimeRouter, { strict: false });
+      expect(router.active.kind).toBe('harness');
+    } finally {
+      await moduleRef.close();
+    }
+  });
+});
+
+/**
+ * Task Five, Step Two — group 2 (WHOLE production `AppModule` boot, legacy end-to-end wiring).
+ *
+ * The groups above bound away the heavy cycle to isolate the router. This group instead boots the
+ * ACTUAL production `AppModule` (the exact graph `main.ts` runs) in the default LEGACY chat-runtime
+ * mode, overriding ONLY the storage/network side-effect adapters so the boot is bounded and offline
+ * — never the chat/router/harness/reload/commands surface under test. The bounded fakes are exactly
+ * the disk/network leaves:
+ *   - `ProviderService` (the #1 hang risk: its real `onModuleInit` starts an unref'd health-check
+ *     `setInterval` and fetches Ollama over HTTP) → inert no-op instance;
+ *   - `DB_HANDLE`/`DB` → a fake Drizzle-shaped handle that satisfies `runPgliteMigrations` (the local
+ *     tier's `DatabaseModule.onModuleInit`) AND `DefaultRoutingRulesSeed.onModuleInit` (which reads a
+ *     system-rule count — the fake reports rules already present so the seed insert is skipped),
+ *     opening no real database;
+ *   - `STORAGE_ADAPTER`/`MEMORY`/`MEMORY_ADAPTER`/`AUTH`/`BRAIN`/`LOG_SERVICE` → inert fakes so no
+ *     storage/auth/log backend is contacted.
+ * Local tier (the repo's `mosaic.config.json`) already disables BullMQ/Redis and the queue handles;
+ * Discord/Telegram/MCP plugins are env-gated and disarmed by deleting their tokens. Nothing about the
+ * router, `ChatModule`, `HarnessModule`, or `ChatGateway` is faked — those come from the REAL graph.
+ *
+ * The boot+init MUST SUCCEED cleanly (proven by `beforeAll` completing and the ChatGateway test
+ * passing). Red-first: on this branch `ChatRuntimeRouter` is registered in NO module (ChatModule
+ * provides only ChatGateway), so `moduleRef.get(ChatRuntimeRouter)` throws `UnknownElementException`
+ * — a WIRING gap, NOT an init failure. That single retrieval is the intended behavioural red; it
+ * flips green once Step Three registers the exclusive router. The ChatGateway retrieval and its
+ * browser-facing method surface are asserted alongside and pass today, pinning that the boot itself
+ * is healthy so the router failure cannot be mistaken for a mis-shaped fake or an unbounded side
+ * effect.
+ */
+describe('AppModule production boot — legacy ChatRuntimeRouter wiring (Task Five, Step Two group 2)', () => {
+  // A Drizzle-shaped fake that satisfies both DB consumers reached during a local-tier init:
+  //   • runPgliteMigrations(): reads handle.db.$client.exec + handle.db.execute(SELECT hashes);
+  //     exec is a no-op and execute yields an empty ledger, so migration statements no-op through.
+  //   • DefaultRoutingRulesSeed.seedDefaultRules(): db.select().from().where() must resolve to a
+  //     row set — we report a non-zero system-rule count so the seeding INSERT branch is skipped.
+  const fakeDb = {
+    $client: { exec: async (): Promise<void> => {} },
+    execute: async (): Promise<{ rows: unknown[] }> => ({ rows: [] }),
+    select: () => ({
+      from: () => ({
+        where: async (): Promise<Array<{ count: number }>> => [{ count: 1 }],
+      }),
+    }),
+    insert: () => ({ values: async (): Promise<void> => {} }),
+  };
+  const fakeDbHandle = { db: fakeDb, close: async (): Promise<void> => {} };
+  const fakeStorageAdapter = {
+    name: 'fake',
+    migrate: async (): Promise<void> => {},
+    close: async (): Promise<void> => {},
+  };
+  // Inert stand-in for the real ProviderService: no health-check interval, no Ollama fetch.
+  const fakeProviderService = {
+    onModuleInit: async (): Promise<void> => {},
+    onModuleDestroy: (): void => {},
+    getRegistry: () => ({
+      getAvailable: () => [],
+      getAll: () => [],
+      find: () => undefined,
+    }),
+    getDefaultModel: () => undefined,
+    listAvailableModels: () => [],
+    listProviders: () => [],
+    getAdapter: () => undefined,
+    getProvidersHealth: () => [],
+  };
+  const fakeBrain = { conversations: {}, agents: {} };
+
+  const BOOT_TIMEOUT_MS = 120_000;
+
+  let moduleRef: TestingModule;
+  let envSnapshot: Record<string, string | undefined>;
+
+  beforeAll(async () => {
+    envSnapshot = { ...process.env };
+    // Env hygiene: disarm the network-facing plugins/adapters and pin the legacy runtime mode.
+    delete process.env['DATABASE_URL'];
+    delete process.env['DISCORD_BOT_TOKEN'];
+    delete process.env['TELEGRAM_BOT_TOKEN'];
+    delete process.env['MCP_SERVERS'];
+    delete process.env['CHAT_HARNESS_RUNTIME']; // resolveChatRuntimeMode → 'legacy'
+    process.env['MOSAIC_STORAGE_TIER'] = 'local';
+
+    moduleRef = await Test.createTestingModule({ imports: [AppModule] })
+      // Storage/network side-effect adapters ONLY — never the router/chat/harness surface under test.
+      .overrideProvider('DB_HANDLE')
+      .useValue(fakeDbHandle)
+      .overrideProvider('DB')
+      .useValue(fakeDb)
+      .overrideProvider('STORAGE_ADAPTER')
+      .useValue(fakeStorageAdapter)
+      .overrideProvider('AUTH')
+      .useValue({})
+      .overrideProvider('BRAIN')
+      .useValue(fakeBrain)
+      .overrideProvider('LOG_SERVICE')
+      .useValue({})
+      .overrideProvider('MEMORY')
+      .useValue({})
+      .overrideProvider('MEMORY_ADAPTER')
+      .useValue({})
+      .overrideProvider(ProviderService)
+      .useValue(fakeProviderService)
+      .compile();
+
+    // The boot itself MUST succeed cleanly — a rejection here is a bounding failure, not the red.
+    await moduleRef.init();
+  }, BOOT_TIMEOUT_MS);
+
+  afterAll(async () => {
+    if (moduleRef) await moduleRef.close();
+    for (const key of Object.keys(process.env)) {
+      if (!(key in envSnapshot)) delete process.env[key];
+    }
+    for (const [key, value] of Object.entries(envSnapshot)) {
+      if (value === undefined) delete process.env[key];
+      else process.env[key] = value;
+    }
+  });
+
+  // Passes TODAY: the real ChatGateway is provided by the real ChatModule and its browser-facing
+  // surface exists. This pins that the whole-AppModule boot came up healthy, so the router failure
+  // below is unambiguously a wiring gap and not a mis-shaped fake or an unbounded side effect.
+  it('boots the whole AppModule and exposes the real ChatGateway with its browser-facing methods', () => {
+    const gateway = moduleRef.get(ChatGateway, { strict: false });
+    expect(typeof gateway.broadcastReload).toBe('function');
+    expect(typeof gateway.getModelOverride).toBe('function');
+    expect(typeof gateway.setModelOverride).toBe('function');
+    expect(typeof gateway.broadcastSessionInfo).toBe('function');
+  });
+
+  // RED TODAY: ChatRuntimeRouter is registered in no module on this branch, so this retrieval throws
+  // UnknownElementException — the intended red-first wiring failure. GREEN once Step Three registers
+  // the exclusive router in the production graph, where legacy mode resolves the embedded runtime.
+  it('resolves the exclusive ChatRuntimeRouter to the embedded runtime in legacy mode', () => {
+    const router = moduleRef.get(ChatRuntimeRouter, { strict: false });
+    expect(router.active.kind).toBe('embedded');
+  });
+});

apps/gateway/src/chat/chat-runtime-router.ts

@@ -0,0 +1,173 @@
+import { Injectable, type OnModuleInit } from '@nestjs/common';
+import { HarnessRegistry } from '../harness/harness.registry.js';
+import {
+  isHarnessConversationServiceAvailable,
+  type HarnessConversationServiceBinding,
+} from '../harness/harness.tokens.js';
+import type {
+  ChatRuntime,
+  ChatRuntimeMode,
+  LegacyBrowserMessagePayload,
+  LegacyEmbeddedChatPort,
+  LegacyRuntimeResult,
+  LegacyRuntimeStream,
+  LegacySessionPresentation,
+  LegacySocketTurnLease,
+  OwnedConversationContext,
+  VerifiedDiscordIngressContext,
+  VerifiedDiscordTurnLease,
+} from './chat-runtime.js';
+import { ChatRuntimeUnavailableError, resolveChatRuntimeMode } from './chat-runtime.js';
+
+/** The fixed fail-closed result for a legacy browser operation issued under `pi-rpc`. */
+const RUNTIME_UNSUPPORTED = {
+  ok: false as const,
+  code: 'runtime_unsupported' as const,
+  retryable: false as const,
+};
+
+/**
+ * Resolves the one live {@link ChatRuntime} for this process and enforces the
+ * `pi-rpc` readiness preconditions at module init — before the gateway accepts
+ * traffic. It never falls back from `pi-rpc` to embedded execution: an unmet
+ * `pi-rpc` precondition is a typed startup failure ({@link ChatRuntimeUnavailableError}),
+ * and until `onModuleInit` selects a runtime, {@link active} throws rather than
+ * exposing any runtime — a failed `pi-rpc` init can never leak the embedded one.
+ */
+@Injectable()
+export class ChatRuntimeRouter implements OnModuleInit, LegacyEmbeddedChatPort {
+  private readonly mode: ChatRuntimeMode;
+
+  /** The single resolved runtime. Undefined until a successful `onModuleInit`. */
+  private resolved: ChatRuntime | undefined;
+
+  constructor(
+    private readonly harnessRegistry: HarnessRegistry,
+    private readonly conversationService: HarnessConversationServiceBinding,
+    private readonly embedded: ChatRuntime,
+    private readonly harness: ChatRuntime,
+    mode: ChatRuntimeMode = resolveChatRuntimeMode(),
+  ) {
+    this.mode = mode;
+  }
+
+  onModuleInit(): void {
+    if (this.mode === 'legacy') {
+      // Legacy ignores the pi-rpc preconditions entirely and always runs embedded.
+      this.resolved = this.embedded;
+      return;
+    }
+
+    // pi-rpc: both preconditions are hard startup failures, checked in a fixed order.
+    if (!this.harnessRegistry.has('pi')) {
+      this.resolved = undefined;
+      throw new ChatRuntimeUnavailableError('adapter_unavailable');
+    }
+    if (!isHarnessConversationServiceAvailable(this.conversationService)) {
+      this.resolved = undefined;
+      throw new ChatRuntimeUnavailableError('conversation_service_unavailable');
+    }
+
+    this.resolved = this.harness;
+  }
+
+  get active(): ChatRuntime {
+    if (this.resolved === undefined) {
+      // Reached only if init has not run or failed closed; never expose a runtime here.
+      throw new Error('The chat runtime is not available: startup did not resolve a runtime.');
+    }
+    return this.resolved;
+  }
+
+  /**
+   * The process-wide mode, available before {@link onModuleInit}. Production handlers read
+   * this to fail a legacy browser turn closed under `pi-rpc` *before* parsing the payload as
+   * either browser-legacy input or a Discord envelope — never to branch into a fallback.
+   */
+  get runtimeMode(): ChatRuntimeMode {
+    return this.mode;
+  }
+
+  /**
+   * The embedded runtime narrowed to its port. Only reached on the legacy path (and for the
+   * verified-Discord op in both modes), where the injected runtime is always a real
+   * `EmbeddedChatRuntime`. The router spec constructs the router with a bare `{ kind }` stub
+   * but never invokes a port op, so this narrowing is never exercised against the stub.
+   */
+  private get embeddedPort(): LegacyEmbeddedChatPort {
+    return this.embedded as unknown as LegacyEmbeddedChatPort;
+  }
+
+  // --- LegacyEmbeddedChatPort: legacy browser operations fail closed under pi-rpc ---
+
+  completeLegacyRestTurn(
+    context: OwnedConversationContext,
+    input: Readonly<{ content: string }>,
+  ): Promise<
+    LegacyRuntimeResult<Readonly<{ text: string; presentation: LegacySessionPresentation }>>
+  > {
+    if (this.mode === 'pi-rpc') {
+      return Promise.resolve(RUNTIME_UNSUPPORTED);
+    }
+    return this.embeddedPort.completeLegacyRestTurn(context, input);
+  }
+
+  prepareLegacySocketTurn(
+    context: OwnedConversationContext,
+    input: LegacyBrowserMessagePayload,
+    stream: LegacyRuntimeStream,
+  ): Promise<LegacyRuntimeResult<LegacySocketTurnLease>> {
+    if (this.mode === 'pi-rpc') {
+      return Promise.resolve(RUNTIME_UNSUPPORTED);
+    }
+    return this.embeddedPort.prepareLegacySocketTurn(context, input, stream);
+  }
+
+  setLegacyThinking(
+    context: OwnedConversationContext,
+    level: string,
+  ): LegacyRuntimeResult<LegacySessionPresentation> {
+    if (this.mode === 'pi-rpc') {
+      return RUNTIME_UNSUPPORTED;
+    }
+    return this.embeddedPort.setLegacyThinking(context, level);
+  }
+
+  abortLegacyTurn(context: OwnedConversationContext): Promise<LegacyRuntimeResult<void>> {
+    if (this.mode === 'pi-rpc') {
+      return Promise.resolve(RUNTIME_UNSUPPORTED);
+    }
+    return this.embeddedPort.abortLegacyTurn(context);
+  }
+
+  applyLegacyModelOverride(
+    context: OwnedConversationContext,
+    modelId: string,
+  ): LegacyRuntimeResult<LegacySessionPresentation> {
+    if (this.mode === 'pi-rpc') {
+      return RUNTIME_UNSUPPORTED;
+    }
+    return this.embeddedPort.applyLegacyModelOverride(context, modelId);
+  }
+
+  readLegacySessionPresentation(
+    context: OwnedConversationContext,
+  ): LegacyRuntimeResult<LegacySessionPresentation> {
+    if (this.mode === 'pi-rpc') {
+      return RUNTIME_UNSUPPORTED;
+    }
+    return this.embeddedPort.readLegacySessionPresentation(context);
+  }
+
+  /**
+   * Verified Discord ingress bypasses browser mode: it is embedded-only in BOTH modes and
+   * never reaches the harness or routing-engine selection. It is reached only through a
+   * {@link VerifiedDiscordIngressContext}, which exists only after every ingress check.
+   */
+  dispatchVerifiedDiscordIngress(
+    context: VerifiedDiscordIngressContext,
+    stream: LegacyRuntimeStream,
+  ): Promise<LegacyRuntimeResult<VerifiedDiscordTurnLease>> {
+    return this.embeddedPort.dispatchVerifiedDiscordIngress(context, stream);
+  }
+}

apps/gateway/src/chat/chat-runtime.ts

@@ -0,0 +1,273 @@
+import type { ChannelAttachmentDto, RoutingDecisionInfo } from '@mosaicstack/types';
+
+/**
+ * The single chat execution strategy resolved by {@link ChatRuntimeRouter}.
+ *
+ * Exactly one runtime is live per process. There is no union that lets a
+ * `pi-rpc` deployment silently fall back to embedded execution: an unmet
+ * `pi-rpc` precondition is a typed startup failure, never a downgrade.
+ */
+export type ChatRuntimeMode = 'legacy' | 'pi-rpc';
+
+export type ChatRuntimeKind = 'embedded' | 'harness';
+
+/** The resolved runtime. Slice Zero exposes only its immutable {@link ChatRuntimeKind}. */
+export interface ChatRuntime {
+  readonly kind: ChatRuntimeKind;
+}
+
+/** Why the `pi-rpc` runtime could not be made ready. Both are hard startup failures. */
+export type ChatRuntimeUnavailableReason =
+  | 'adapter_unavailable'
+  | 'conversation_service_unavailable';
+
+/**
+ * Raised at module init when `pi-rpc` mode is selected but its preconditions are
+ * unmet. Carries only fixed, browser-safe text — never a raw exception message,
+ * stack, or provider detail — and reports the frozen ack code `runtime_unsupported`.
+ */
+export class ChatRuntimeUnavailableError extends Error {
+  readonly code = 'runtime_unsupported' as const;
+  readonly reason: ChatRuntimeUnavailableReason;
+
+  constructor(reason: ChatRuntimeUnavailableReason) {
+    super(
+      reason === 'adapter_unavailable'
+        ? 'The pi-rpc chat runtime is unavailable: no "pi" harness adapter is registered.'
+        : 'The pi-rpc chat runtime is unavailable: the harness conversation service is not bound.',
+    );
+    this.name = 'ChatRuntimeUnavailableError';
+    this.reason = reason;
+  }
+}
+
+/**
+ * Resolves the process-wide chat runtime mode from the environment. Anything other
+ * than the exact opt-in token `pi-rpc` keeps the legacy embedded runtime.
+ */
+export function resolveChatRuntimeMode(
+  env: Record<string, string | undefined> = process.env,
+): ChatRuntimeMode {
+  return env['CHAT_HARNESS_RUNTIME'] === 'pi-rpc' ? 'pi-rpc' : 'legacy';
+}
+
+// ---------------------------------------------------------------------------
+// Transitional embedded chat port (Task Five).
+//
+// The legacy embedded browser behaviour is moved behind this exact interface so
+// neither the controller nor the gateway retains AgentService, RoutingEngine,
+// session, `piSession`, metric, listener, or channel access. `EmbeddedChatRuntime`
+// implements the port; `ChatRuntimeRouter` exposes the same narrowly named
+// operations and returns `runtime_unsupported` before touching Embedded for legacy
+// browser operations when the mode is `pi-rpc`.
+//
+// The names are frozen (spec jarvis-brain@1c629b06). Legacy REST completion,
+// legacy Socket streaming, P3 harness turns, and verified Discord are distinct
+// transport/trust capabilities — there is deliberately no generic
+// `sendConversationTurn` nor an AgentService-shaped mirror on the router.
+// ---------------------------------------------------------------------------
+
+/**
+ * Phantom brand keeping {@link OwnedConversationContext} nominally distinct so browser
+ * DTOs are never structurally assignable to it. The factory that mints one may be called
+ * only after authentication with `scopeFromUser(...)`, never with payload authority fields.
+ */
+declare const ownedConversationContextBrand: unique symbol;
+
+/** Gateway-only ownership context. Embedded rechecks owner+tenant on every operation. */
+export interface OwnedConversationContext {
+  readonly [ownedConversationContextBrand]: true;
+  readonly conversationId: string;
+  readonly scope: Readonly<{ userId: string; tenantId: string }>;
+}
+
+/**
+ * Every non-`ok` legacy runtime outcome. Missing, foreign, and no-longer-owned
+ * conversations all collapse to `conversation_unavailable`. Ownership/mode/validation
+ * failures are total results and never throw.
+ */
+export type LegacyRuntimeFailure =
+  | { readonly ok: false; readonly code: 'runtime_unsupported'; readonly retryable: false }
+  | { readonly ok: false; readonly code: 'conversation_unavailable'; readonly retryable: false }
+  | { readonly ok: false; readonly code: 'request_invalid'; readonly retryable: false }
+  | {
+      readonly ok: false;
+      readonly code: 'thinking_level_invalid';
+      readonly retryable: false;
+      readonly availableThinkingLevels: readonly string[];
+    }
+  | { readonly ok: false; readonly code: 'runtime_unavailable'; readonly retryable: true }
+  | { readonly ok: false; readonly code: 'turn_already_dispatched'; readonly retryable: false }
+  | { readonly ok: false; readonly code: 'operation_failed'; readonly retryable: boolean }
+  | { readonly ok: false; readonly code: 'timeout'; readonly retryable: true };
+
+/** Total result: an `ok` value or one of the fixed {@link LegacyRuntimeFailure} codes. */
+export type LegacyRuntimeResult<T> =
+  | { readonly ok: true; readonly value: T }
+  | LegacyRuntimeFailure;
+
+/** User-facing session projection. Carries no session object, handle, or credential path. */
+export interface LegacySessionPresentation {
+  readonly provider: string;
+  readonly modelId: string;
+  readonly thinkingLevel: string;
+  readonly availableThinkingLevels: readonly string[];
+  readonly agentName?: string;
+  readonly routingDecision?: RoutingDecisionInfo;
+}
+
+/** Terminal usage stats, normalized by Embedded from AgentService metrics. */
+export interface LegacyUsage {
+  readonly provider: string;
+  readonly modelId: string;
+  readonly thinkingLevel: string;
+  readonly tokens: Readonly<{
+    input: number;
+    output: number;
+    cacheRead: number;
+    cacheWrite: number;
+    total: number;
+  }>;
+  readonly cost: number;
+  readonly context: Readonly<{ percent: number | null; window: number }>;
+}
+
+/**
+ * Normalized stream event. Exposes no `AgentSession`, `piSession`, native handle, raw
+ * exception, tool arguments, or credential-bearing path — the gateway sees only these.
+ */
+export type LegacyRuntimeEvent =
+  | { readonly type: 'started' }
+  | { readonly type: 'text_delta'; readonly text: string }
+  | { readonly type: 'thinking_delta'; readonly text: string }
+  | {
+      readonly type: 'tool_started';
+      readonly toolCallId: string;
+      readonly toolName: string;
+    }
+  | {
+      readonly type: 'tool_finished';
+      readonly toolCallId: string;
+      readonly toolName: string;
+      readonly isError: boolean;
+    }
+  | { readonly type: 'settled'; readonly usage?: LegacyUsage };
+
+/** Legacy browser message input. Authority fields are advisory only; scope comes from the context. */
+export interface LegacyBrowserMessagePayload {
+  readonly content: string;
+  readonly provider?: string;
+  readonly modelId?: string;
+  readonly agentId?: string;
+  readonly attachments?: readonly ChannelAttachmentDto[];
+}
+
+/** A prepared-but-not-yet-dispatched legacy socket turn. */
+export interface LegacySocketTurnLease {
+  readonly presentation: LegacySessionPresentation;
+  /**
+   * Atomically one-shot and scope-rechecking. A second call returns
+   * `turn_already_dispatched` and performs zero prompt/tool effects.
+   */
+  dispatch(): Promise<LegacyRuntimeResult<void>>;
+  /** Idempotent, non-throwing. Removes listener and channel, including partial setup. */
+  dispose(): Promise<void>;
+}
+
+/**
+ * Phantom brand for {@link VerifiedDiscordIngressContext}. Minted only after service-token
+ * auth plus signature, allowlist, binding, expected-route, replay, configured-agent,
+ * forced-scope, and attachment-normalization checks.
+ */
+declare const verifiedDiscordIngressContextBrand: unique symbol;
+
+/** Fully-verified Discord ingress. Contains no socket, envelope, signature, token, or escape hatch. */
+export interface VerifiedDiscordIngressContext {
+  readonly [verifiedDiscordIngressContextBrand]: true;
+  readonly conversationId: string;
+  readonly scope: Readonly<{ userId: string; tenantId: string }>;
+  readonly configuredAgent: Readonly<{ agentConfigId: string; instanceId: string }>;
+  readonly content: string;
+  readonly attachments?: readonly ChannelAttachmentDto[];
+  readonly correlationId: string;
+  readonly discordMessageId: string;
+  readonly discordUserId: string;
+}
+
+/** Verified-Discord turn lease. Same atomic one-shot dispatch and idempotent dispose rules. */
+export interface VerifiedDiscordTurnLease {
+  readonly presentation: LegacySessionPresentation;
+  dispatch(): Promise<LegacyRuntimeResult<void>>;
+  dispose(): Promise<void>;
+}
+
+/** Server-owned egress projection the runtime pushes normalized events into. */
+export interface LegacyRuntimeStream {
+  /** Server-derived, e.g. `websocket:<socket-id>`. Never client-supplied. */
+  readonly channelId: string;
+  onEvent(event: LegacyRuntimeEvent): void;
+}
+
+/**
+ * The exact transitional port. `EmbeddedChatRuntime` implements it; `ChatRuntimeRouter`
+ * mirrors the operation names and fails closed with `runtime_unsupported` for legacy
+ * browser operations under `pi-rpc`.
+ */
+export interface LegacyEmbeddedChatPort {
+  completeLegacyRestTurn(
+    context: OwnedConversationContext,
+    input: Readonly<{ content: string }>,
+  ): Promise<
+    LegacyRuntimeResult<Readonly<{ text: string; presentation: LegacySessionPresentation }>>
+  >;
+
+  prepareLegacySocketTurn(
+    context: OwnedConversationContext,
+    input: LegacyBrowserMessagePayload,
+    stream: LegacyRuntimeStream,
+  ): Promise<LegacyRuntimeResult<LegacySocketTurnLease>>;
+
+  setLegacyThinking(
+    context: OwnedConversationContext,
+    level: string,
+  ): LegacyRuntimeResult<LegacySessionPresentation>;
+
+  abortLegacyTurn(context: OwnedConversationContext): Promise<LegacyRuntimeResult<void>>;
+
+  applyLegacyModelOverride(
+    context: OwnedConversationContext,
+    modelId: string,
+  ): LegacyRuntimeResult<LegacySessionPresentation>;
+
+  readLegacySessionPresentation(
+    context: OwnedConversationContext,
+  ): LegacyRuntimeResult<LegacySessionPresentation>;
+
+  dispatchVerifiedDiscordIngress(
+    context: VerifiedDiscordIngressContext,
+    stream: LegacyRuntimeStream,
+  ): Promise<LegacyRuntimeResult<VerifiedDiscordTurnLease>>;
+}
+
+/**
+ * Mints an {@link OwnedConversationContext} from a server-derived scope. Callers must pass
+ * a scope produced by `scopeFromUser(...)` after authentication — never a client-supplied
+ * authority field. The brand is phantom, so this is the only way to obtain the branded type.
+ */
+export function ownConversation(
+  conversationId: string,
+  scope: Readonly<{ userId: string; tenantId: string }>,
+): OwnedConversationContext {
+  return { conversationId, scope } as unknown as OwnedConversationContext;
+}
+
+/**
+ * Mints a {@link VerifiedDiscordIngressContext}. Callers must have already completed every
+ * ingress check (service-token auth, signature, allowlist, binding, expected-route, replay,
+ * configured-agent, forced-scope, attachment normalization) before calling this.
+ */
+export function verifyDiscordIngress(
+  fields: Omit<VerifiedDiscordIngressContext, typeof verifiedDiscordIngressContextBrand>,
+): VerifiedDiscordIngressContext {
+  return { ...fields } as unknown as VerifiedDiscordIngressContext;
+}

apps/gateway/src/chat/chat.controller.ts

@@ -3,21 +3,20 @@ import {
   Post,
   Body,
   Logger,
-  ForbiddenException,
   HttpException,
   HttpStatus,
   NotFoundException,
-  Inject,
   UseGuards,
 } from '@nestjs/common';
-import type { AgentSessionEvent } from '@mariozechner/pi-coding-agent';
 import { Throttle } from '@nestjs/throttler';
-import { AgentService } from '../agent/agent.service.js';
 import { AuthGuard } from '../auth/auth.guard.js';
 import { CurrentUser } from '../auth/current-user.decorator.js';
 import { scopeFromUser, type AuthenticatedUserLike } from '../auth/session-scope.js';
 import { v4 as uuid } from 'uuid';
 import { ChatRequestDto } from './chat.dto.js';
+import { ChatRuntimeRouter } from './chat-runtime-router.js';
+import { ownConversation } from './chat-runtime.js';
+import type { LegacyRuntimeFailure } from './chat-runtime.js';
 
 interface ChatResponse {
   conversationId: string;
@@ -29,7 +28,7 @@ interface ChatResponse {
 export class ChatController {
   private readonly logger = new Logger(ChatController.name);
 
-  constructor(@Inject(AgentService) private readonly agentService: AgentService) {}
+  constructor(private readonly runtime: ChatRuntimeRouter) {}
 
   @Post()
   @Throttle({ default: { limit: 10, ttl: 60_000 } })
@@ -40,68 +39,38 @@ export class ChatController {
     const conversationId = body.conversationId ?? uuid();
     const scope = scopeFromUser(user);
 
-    try {
-      let agentSession = this.agentService.getSession(conversationId, scope);
-      if (!agentSession) {
-        agentSession = await this.agentService.createSession(conversationId, {
-          userId: scope.userId,
-          tenantId: scope.tenantId,
-        });
-      }
-    } catch (err) {
-      if (err instanceof ForbiddenException) {
-        throw new NotFoundException('Session not found');
-      }
-      this.logger.error(
-        `Session creation failed for conversation=${conversationId}`,
-        err instanceof Error ? err.stack : String(err),
-      );
-      throw new HttpException('Agent session unavailable', HttpStatus.SERVICE_UNAVAILABLE);
-    }
-
     this.logger.debug(`Handling chat request for user=${user.id}, conversation=${conversationId}`);
 
-    let responseText = '';
+    // The one exclusive runtime owns execution. In legacy mode this reaches the embedded runtime;
+    // in pi-rpc it fails closed with `runtime_unsupported` before ever touching embedded execution.
+    const result = await this.runtime.completeLegacyRestTurn(
+      ownConversation(conversationId, scope),
+      { content: body.content },
+    );
 
-    const done = new Promise<void>((resolve, reject) => {
+    if (result.ok) {
-      const timer = setTimeout(() => {
+      return { conversationId, text: result.value.text };
-        cleanup();
-        this.logger.error(`Agent response timed out after 120s for conversation=${conversationId}`);
-        reject(new Error('Agent response timed out'));
-      }, 120_000);
-
-      const cleanup = this.agentService.onEvent(
-        conversationId,
-        (event: AgentSessionEvent) => {
-          if (
-            event.type === 'message_update' &&
-            event.assistantMessageEvent.type === 'text_delta'
-          ) {
-            responseText += event.assistantMessageEvent.delta;
-          }
-          if (event.type === 'agent_end') {
-            clearTimeout(timer);
-            cleanup();
-            resolve();
-          }
-        },
-        scope,
-      );
-    });
-
-    try {
-      await this.agentService.prompt(conversationId, body.content, scope);
-      await done;
-    } catch (err) {
-      if (err instanceof HttpException) throw err;
-      const message = err instanceof Error ? err.message : String(err);
-      if (message.includes('timed out')) {
-        throw new HttpException('Agent response timed out', HttpStatus.GATEWAY_TIMEOUT);
-      }
-      this.logger.error(`Chat prompt failed for conversation=${conversationId}`, String(err));
-      throw new HttpException('Agent processing failed', HttpStatus.INTERNAL_SERVER_ERROR);
     }
 
-    return { conversationId, text: responseText };
+    throw this.toHttpException(result, conversationId);
+  }
+
+  /** Maps a total {@link LegacyRuntimeFailure} to the fixed browser-safe HTTP surface. */
+  private toHttpException(failure: LegacyRuntimeFailure, conversationId: string): HttpException {
+    switch (failure.code) {
+      case 'conversation_unavailable':
+        return new NotFoundException('Session not found');
+      case 'request_invalid':
+      case 'thinking_level_invalid':
+        return new HttpException('Invalid chat request', HttpStatus.BAD_REQUEST);
+      case 'timeout':
+        return new HttpException('Agent response timed out', HttpStatus.GATEWAY_TIMEOUT);
+      case 'runtime_unsupported':
+      case 'runtime_unavailable':
+        return new HttpException('Agent runtime unavailable', HttpStatus.SERVICE_UNAVAILABLE);
+      default:
+        this.logger.error(`Chat turn failed for conversation=${conversationId}: ${failure.code}`);
+        return new HttpException('Agent processing failed', HttpStatus.INTERNAL_SERVER_ERROR);
+    }
   }
 }

apps/gateway/src/chat/chat.dto.ts

@@ -1,5 +1,14 @@
 import type { ChannelAttachmentDto } from '@mosaicstack/types';
-import { IsOptional, IsString, IsUUID, MaxLength } from 'class-validator';
+import { Transform, Type } from 'class-transformer';
+import {
+  IsNotEmpty,
+  IsObject,
+  IsOptional,
+  IsString,
+  IsUUID,
+  MaxLength,
+  ValidateNested,
+} from 'class-validator';
 
 export class ChatRequestDto {
   @IsOptional()
@@ -37,3 +46,56 @@ export class ChatSocketMessageDto {
   /** Validated channel attachment references; binary content is not embedded. */
   attachments?: readonly ChannelAttachmentDto[];
 }
+
+/**
+ * Task Five, group 2 — the frozen pi-rpc `turn:send` selection triple.
+ *
+ * Each id is a required, non-empty, bounded string. There is no `@IsOptional` and no extra
+ * field: under `forbidNonWhitelisted` an unknown selection key is rejected, and a missing id
+ * fails `@IsString` (undefined is not a string) rather than silently passing.
+ */
+export class HarnessTurnSelectionDto {
+  @IsString()
+  @IsNotEmpty()
+  @MaxLength(255)
+  harnessId!: string;
+
+  @IsString()
+  @IsNotEmpty()
+  @MaxLength(255)
+  providerId!: string;
+
+  @IsString()
+  @IsNotEmpty()
+  @MaxLength(255)
+  modelId!: string;
+}
+
+/**
+ * Task Five, group 2 — the frozen wire contract for a pi-rpc `turn:send`.
+ *
+ * Validated through the production `ValidationPipe({ whitelist, forbidNonWhitelisted, transform })`:
+ * a UUID conversation id; `content` trimmed then bounded to 1..10_000 characters (whitespace-only
+ * collapses to empty and fails `@IsNotEmpty`); a nested `selection` object recursed with an
+ * explicit `@Type` (a bare `@ValidateNested` is masked green by class-validator's empty-metadata
+ * `unknownValue`); and a UUID-v4 idempotency key. No `provider`/`modelId`/`attachments` or other
+ * authority field is declared, so `forbidNonWhitelisted` rejects every unknown top-level key.
+ */
+export class HarnessTurnSendDto {
+  @IsUUID()
+  conversationId!: string;
+
+  @Transform(({ value }) => (typeof value === 'string' ? value.trim() : value))
+  @IsString()
+  @IsNotEmpty()
+  @MaxLength(10_000)
+  content!: string;
+
+  @IsObject()
+  @ValidateNested()
+  @Type(() => HarnessTurnSelectionDto)
+  selection!: HarnessTurnSelectionDto;
+
+  @IsUUID('4')
+  idempotencyKey!: string;
+}

apps/gateway/src/chat/chat.gateway-command-approval.spec.ts

@@ -8,12 +8,31 @@ const payload: SlashCommandPayload = {
   approvalId: 'approval-1',
 };
 
+/**
+ * Task 5 fence (F, existing control): gateway-owned command authorization/approval must
+ * cause ZERO chat-runtime dispatch. Placed in the gateway's chat-runtime-router slot (the
+ * former direct `AgentService` slot) so any accidental chat-runtime resolution throws
+ * loudly instead of silently passing. Because execute/approval run entirely through the
+ * command executor dependency and never resolve a chat runtime, this fixture is never
+ * triggered and the ingress stays a GREEN control.
+ */
+function failIfUsedChatRuntimeRouter() {
+  return {
+    onModuleInit: () => {
+      throw new Error('chat runtime router must not initialise on the command approval path');
+    },
+    get active(): never {
+      throw new Error('chat runtime must not be resolved on the command approval path');
+    },
+  };
+}
+
 function buildGateway(commandExecutor: {
   execute: ReturnType<typeof vi.fn>;
   createApproval: ReturnType<typeof vi.fn>;
 }): ChatGateway {
   return new ChatGateway(
-    {} as never,
+    failIfUsedChatRuntimeRouter() as never,
     {} as never,
     {} as never,
     {} as never,
@@ -72,3 +91,114 @@ describe('ChatGateway command approval ingress', () => {
     });
   });
 });
+
+/**
+ * Task 5 (G3) command runtime fence. Under pi-rpc there is no embedded chat session, so
+ * embedded slash-commands (/model, /agent, and every other non-audited command) are fixed
+ * "unsupported" and MUST fail closed BEFORE reaching the command executor — never a silent
+ * fall-through to embedded execution. Only runtime-independent audited system commands
+ * (/reload) pass through as a positive control, and the approval path stays runtime-independent.
+ * The router stub here carries `runtimeMode: 'pi-rpc'` and throws if any runtime is resolved, so
+ * a fence bypass surfaces as a thrown error rather than a silent embedded dispatch.
+ */
+function buildPiRpcGateway(commandExecutor: {
+  execute: ReturnType<typeof vi.fn>;
+  createApproval: ReturnType<typeof vi.fn>;
+}): ChatGateway {
+  const piRpcRouter = {
+    runtimeMode: 'pi-rpc' as const,
+    onModuleInit: () => {
+      throw new Error('chat runtime router must not initialise on the pi-rpc command path');
+    },
+    get active(): never {
+      throw new Error('chat runtime must not be resolved on the pi-rpc command path');
+    },
+  };
+  return new ChatGateway(
+    piRpcRouter as never,
+    {} as never,
+    {} as never,
+    {} as never,
+    commandExecutor as never,
+    {} as never,
+  );
+}
+
+describe('ChatGateway command runtime fence (Task 5 G3, pi-rpc)', () => {
+  const UNSUPPORTED = 'Slash commands are not available on this deployment.';
+
+  it.each(['model', 'agent', 'gc'])(
+    'fails /%s closed before the executor under pi-rpc (execute never called)',
+    async (command): Promise<void> => {
+      const commandExecutor = {
+        execute: vi
+          .fn()
+          .mockResolvedValue({ command, conversationId: 'conversation-1', success: true }),
+        createApproval: vi.fn(),
+      };
+      const gateway = buildPiRpcGateway(commandExecutor);
+      const client = { data: { user: { id: 'admin-1' } }, emit: vi.fn() };
+
+      await gateway.handleCommandExecute(client as never, {
+        command,
+        conversationId: 'conversation-1',
+      });
+
+      expect(commandExecutor.execute).toHaveBeenCalledTimes(0);
+      expect(client.emit).toHaveBeenCalledWith('command:result', {
+        command,
+        conversationId: 'conversation-1',
+        success: false,
+        message: UNSUPPORTED,
+      });
+    },
+  );
+
+  it('passes the audited /reload system command through as a positive control under pi-rpc', async (): Promise<void> => {
+    const reloadResult = { command: 'reload', conversationId: 'conversation-1', success: true };
+    const commandExecutor = {
+      execute: vi.fn().mockResolvedValue(reloadResult),
+      createApproval: vi.fn(),
+    };
+    const gateway = buildPiRpcGateway(commandExecutor);
+    const client = { data: { user: { id: 'admin-1' } }, emit: vi.fn() };
+
+    await gateway.handleCommandExecute(client as never, {
+      command: 'reload',
+      conversationId: 'conversation-1',
+    });
+
+    expect(commandExecutor.execute).toHaveBeenCalledTimes(1);
+    expect(commandExecutor.execute).toHaveBeenCalledWith(
+      { command: 'reload', conversationId: 'conversation-1' },
+      { userId: 'admin-1', tenantId: 'admin-1' },
+    );
+    expect(client.emit).toHaveBeenCalledWith('command:result', reloadResult);
+  });
+
+  it('keeps command approval runtime-independent under pi-rpc (createApproval still runs)', async (): Promise<void> => {
+    const commandExecutor = {
+      execute: vi.fn(),
+      createApproval: vi.fn().mockResolvedValue({
+        approvalId: 'approval-1',
+        expiresAt: '2026-07-12T00:05:00.000Z',
+      }),
+    };
+    const gateway = buildPiRpcGateway(commandExecutor);
+    const client = { data: { user: { id: 'admin-1' } }, emit: vi.fn() };
+
+    await gateway.handleCommandApproval(client as never, {
+      command: 'gc',
+      conversationId: 'conversation-1',
+    });
+
+    expect(commandExecutor.createApproval).toHaveBeenCalledWith(
+      { command: 'gc', conversationId: 'conversation-1' },
+      { userId: 'admin-1', tenantId: 'admin-1' },
+    );
+    expect(client.emit).toHaveBeenCalledWith(
+      'command:approval',
+      expect.objectContaining({ success: true, approvalId: 'approval-1' }),
+    );
+  });
+});

apps/gateway/src/chat/chat.module.ts

@@ -1,12 +1,59 @@
 import { forwardRef, Module } from '@nestjs/common';
 import { CommandsModule } from '../commands/commands.module.js';
+import { HarnessModule } from '../harness/harness.module.js';
+import { HarnessRegistry } from '../harness/harness.registry.js';
+import {
+  HARNESS_CONVERSATION_SERVICE,
+  HARNESS_REGISTRY,
+  type HarnessConversationServiceBinding,
+} from '../harness/harness.tokens.js';
+import type { HarnessConversationService } from '@mosaicstack/types';
 import { ChatGateway } from './chat.gateway.js';
 import { ChatController } from './chat.controller.js';
+import { ChatRuntimeRouter } from './chat-runtime-router.js';
+import { EmbeddedChatRuntime } from './embedded-chat.runtime.js';
+import { HarnessChatRuntime } from './harness-chat.runtime.js';
 
+/**
+ * Task Five wiring. The exclusive {@link ChatRuntimeRouter} is the single chat-execution
+ * authority: the controller and gateway inject only the router, never `AgentService`,
+ * `RoutingEngineService`, or a session/`piSession` handle. The router resolves exactly one
+ * runtime at module init — {@link EmbeddedChatRuntime} in legacy mode, {@link HarnessChatRuntime}
+ * in `pi-rpc` — over the REAL {@link HarnessModule} registry and conversation-service binding.
+ *
+ * The router and the harness runtime are constructed through factories because their
+ * dependencies are interface/union types with no runtime injection token (the registry and
+ * conversation-service arrive via the string tokens exported by `HarnessModule`); the embedded
+ * runtime injects the class-typed `AgentService` and is provided directly.
+ */
 @Module({
-  imports: [forwardRef(() => CommandsModule)],
+  imports: [forwardRef(() => CommandsModule), HarnessModule],
   controllers: [ChatController],
-  providers: [ChatGateway],
+  providers: [
-  exports: [ChatGateway],
+    ChatGateway,
+    EmbeddedChatRuntime,
+    {
+      provide: HarnessChatRuntime,
+      useFactory: (conversationService: HarnessConversationServiceBinding) =>
+        new HarnessChatRuntime(conversationService as HarnessConversationService),
+      inject: [HARNESS_CONVERSATION_SERVICE],
+    },
+    {
+      provide: ChatRuntimeRouter,
+      useFactory: (
+        registry: HarnessRegistry,
+        conversationService: HarnessConversationServiceBinding,
+        embedded: EmbeddedChatRuntime,
+        harness: HarnessChatRuntime,
+      ) => new ChatRuntimeRouter(registry, conversationService, embedded, harness),
+      inject: [
+        HARNESS_REGISTRY,
+        HARNESS_CONVERSATION_SERVICE,
+        EmbeddedChatRuntime,
+        HarnessChatRuntime,
+      ],
+    },
+  ],
+  exports: [ChatGateway, ChatRuntimeRouter],
 })
 export class ChatModule {}

apps/gateway/src/chat/embedded-chat.runtime.ts

@@ -0,0 +1,532 @@
+import { ForbiddenException, Injectable, Logger, NotFoundException } from '@nestjs/common';
+import type { AgentSessionEvent } from '@mariozechner/pi-coding-agent';
+import { AgentService, type AgentSession } from '../agent/agent.service.js';
+import type { ActorTenantScope } from '../auth/session-scope.js';
+import type {
+  ChatRuntime,
+  LegacyBrowserMessagePayload,
+  LegacyEmbeddedChatPort,
+  LegacyRuntimeEvent,
+  LegacyRuntimeResult,
+  LegacySessionPresentation,
+  LegacySocketTurnLease,
+  LegacyUsage,
+  OwnedConversationContext,
+  VerifiedDiscordIngressContext,
+  VerifiedDiscordTurnLease,
+  LegacyRuntimeStream,
+} from './chat-runtime.js';
+
+/** Fixed timeout for a synchronous REST turn, matching the historical controller budget. */
+const REST_TURN_TIMEOUT_MS = 120_000;
+
+/**
+ * The `legacy` chat runtime and the sole implementation of {@link LegacyEmbeddedChatPort}.
+ *
+ * It owns the embedded in-process execution path — the `AgentService` stack that the
+ * `ChatController` and `ChatGateway` drove directly before Task Five. Once the
+ * {@link import('./chat-runtime-router.js').ChatRuntimeRouter} fronts it, the browser
+ * HTTP/WebSocket legacy path and verified-Discord ingress route through THIS runtime, so
+ * neither the controller nor the gateway retains `AgentService`, `piSession`, session,
+ * listener, channel, or metric access. Ownership (`userId`/`tenantId`) is re-checked by
+ * `AgentService` on every operation; a missing, foreign, or no-longer-owned conversation
+ * collapses to `conversation_unavailable` and never throws out of the port.
+ */
+@Injectable()
+export class EmbeddedChatRuntime implements ChatRuntime, LegacyEmbeddedChatPort {
+  readonly kind = 'embedded' as const;
+  private readonly logger = new Logger(EmbeddedChatRuntime.name);
+
+  constructor(readonly agentService: AgentService) {}
+
+  // -------------------------------------------------------------------------
+  // Legacy REST completion (op A)
+  // -------------------------------------------------------------------------
+
+  async completeLegacyRestTurn(
+    context: OwnedConversationContext,
+    input: Readonly<{ content: string }>,
+  ): Promise<
+    LegacyRuntimeResult<Readonly<{ text: string; presentation: LegacySessionPresentation }>>
+  > {
+    const scope = toScope(context.scope);
+    const { conversationId } = context;
+
+    const resolved = await this.resolveOrCreate(conversationId, scope, {});
+    if (!resolved.ok) return resolved;
+
+    let responseText = '';
+    let timer: ReturnType<typeof setTimeout> | undefined;
+    let detach: (() => void) | undefined;
+    let disposed = false;
+    // One idempotent teardown owned OUTSIDE the completion promise: it clears the timeout and
+    // detaches the event listener exactly once, whichever of agent_end, timeout, or a prompt
+    // rejection fires first. Without this, a prompt() rejection surfaced through the catch below
+    // would return while leaving the listener attached (free to consume a later turn's events) and
+    // the 120s timer live (its rejection later going unobserved).
+    const dispose = (): void => {
+      if (disposed) return;
+      disposed = true;
+      if (timer !== undefined) clearTimeout(timer);
+      detach?.();
+    };
+    const done = new Promise<void>((resolve, reject) => {
+      timer = setTimeout(() => {
+        dispose();
+        reject(new Error('Agent response timed out'));
+      }, REST_TURN_TIMEOUT_MS);
+
+      detach = this.agentService.onEvent(
+        conversationId,
+        (event: AgentSessionEvent) => {
+          if (
+            event.type === 'message_update' &&
+            event.assistantMessageEvent.type === 'text_delta'
+          ) {
+            responseText += event.assistantMessageEvent.delta;
+          }
+          if (event.type === 'agent_end') {
+            dispose();
+            resolve();
+          }
+        },
+        scope,
+      );
+    });
+
+    // Attach the prompt and the completion promise CONCURRENTLY. Awaiting prompt() first left the
+    // timeout unobservable until prompt settled (a hung prompt could never time out) and, worse,
+    // let the 120s timer reject `done` while nothing yet awaited it — a transient unhandledRejection
+    // window. Promise.all installs handlers on BOTH synchronously, so the timeout bounds the whole
+    // turn even while prompt is pending, and neither promise can reject unobserved. Success still
+    // requires both prompt() to resolve AND agent_end to arrive (identical to the prior sequential
+    // await). The idempotent dispose() clears the timer + detaches on whichever settles first.
+    const prompting = this.agentService.prompt(conversationId, input.content, scope);
+    try {
+      await Promise.all([prompting, done]);
+    } catch (err) {
+      dispose();
+      const message = err instanceof Error ? err.message : String(err);
+      if (message.includes('timed out')) {
+        return { ok: false, code: 'timeout', retryable: true };
+      }
+      this.logger.error(`Legacy REST turn failed for conversation=${conversationId}`, message);
+      return { ok: false, code: 'operation_failed', retryable: false };
+    }
+
+    const presentation = this.presentationFor(conversationId, scope) ?? resolved.presentation;
+    return { ok: true, value: { text: responseText, presentation } };
+  }
+
+  // -------------------------------------------------------------------------
+  // Legacy Socket streaming (op B)
+  // -------------------------------------------------------------------------
+
+  async prepareLegacySocketTurn(
+    context: OwnedConversationContext,
+    input: LegacyBrowserMessagePayload,
+    stream: LegacyRuntimeStream,
+  ): Promise<LegacyRuntimeResult<LegacySocketTurnLease>> {
+    const scope = toScope(context.scope);
+    const { conversationId } = context;
+
+    const resolved = await this.resolveOrCreate(conversationId, scope, {
+      ...(input.provider ? { provider: input.provider } : {}),
+      ...(input.modelId ? { modelId: input.modelId } : {}),
+      ...(input.agentId ? { agentConfigId: input.agentId } : {}),
+    });
+    if (!resolved.ok) return resolved;
+
+    let detach: () => void;
+    try {
+      detach = this.subscribe(conversationId, scope, stream);
+    } catch (err) {
+      // A partial listener/channel setup rolled itself back inside subscribe(); surface a total
+      // safe failure instead of throwing out of the port. Retryable — the attach is transient.
+      this.logger.error(
+        `Embedded socket subscription failed for conversation=${conversationId}`,
+        err instanceof Error ? err.message : String(err),
+      );
+      return { ok: false, code: 'runtime_unavailable', retryable: true };
+    }
+
+    return {
+      ok: true,
+      value: this.buildLease(
+        conversationId,
+        scope,
+        input.content,
+        input.attachments,
+        detach,
+        resolved.presentation,
+      ),
+    };
+  }
+
+  // -------------------------------------------------------------------------
+  // Thinking level (op C) — synchronous, total
+  // -------------------------------------------------------------------------
+
+  setLegacyThinking(
+    context: OwnedConversationContext,
+    level: string,
+  ): LegacyRuntimeResult<LegacySessionPresentation> {
+    const scope = toScope(context.scope);
+    const session = this.agentService.getSession(context.conversationId, scope);
+    if (!session) return CONVERSATION_UNAVAILABLE;
+
+    const availableThinkingLevels = session.piSession.getAvailableThinkingLevels();
+    if (!(availableThinkingLevels as readonly string[]).includes(level)) {
+      return {
+        ok: false,
+        code: 'thinking_level_invalid',
+        retryable: false,
+        availableThinkingLevels,
+      };
+    }
+
+    session.piSession.setThinkingLevel(level as never);
+    return { ok: true, value: this.presentationForSession(session) };
+  }
+
+  // -------------------------------------------------------------------------
+  // Abort (op D)
+  // -------------------------------------------------------------------------
+
+  async abortLegacyTurn(context: OwnedConversationContext): Promise<LegacyRuntimeResult<void>> {
+    const scope = toScope(context.scope);
+    const session = this.agentService.getSession(context.conversationId, scope);
+    if (!session) return CONVERSATION_UNAVAILABLE;
+
+    try {
+      await session.piSession.abort();
+    } catch (err) {
+      this.logger.error(
+        `Legacy abort failed for conversation=${context.conversationId}`,
+        err instanceof Error ? err.message : String(err),
+      );
+      return { ok: false, code: 'operation_failed', retryable: false };
+    }
+    return { ok: true, value: undefined };
+  }
+
+  // -------------------------------------------------------------------------
+  // Model override (synchronous, total)
+  // -------------------------------------------------------------------------
+
+  applyLegacyModelOverride(
+    context: OwnedConversationContext,
+    modelId: string,
+  ): LegacyRuntimeResult<LegacySessionPresentation> {
+    const scope = toScope(context.scope);
+    const session = this.agentService.getSession(context.conversationId, scope);
+    if (!session) return CONVERSATION_UNAVAILABLE;
+
+    this.agentService.updateSessionModel(context.conversationId, modelId, scope);
+    const refreshed = this.agentService.getSession(context.conversationId, scope) ?? session;
+    return { ok: true, value: this.presentationForSession(refreshed) };
+  }
+
+  // -------------------------------------------------------------------------
+  // Presentation read (synchronous, total)
+  // -------------------------------------------------------------------------
+
+  readLegacySessionPresentation(
+    context: OwnedConversationContext,
+  ): LegacyRuntimeResult<LegacySessionPresentation> {
+    const scope = toScope(context.scope);
+    const session = this.agentService.getSession(context.conversationId, scope);
+    if (!session) return CONVERSATION_UNAVAILABLE;
+    return { ok: true, value: this.presentationForSession(session) };
+  }
+
+  // -------------------------------------------------------------------------
+  // Verified Discord ingress (embedded-only in both modes)
+  // -------------------------------------------------------------------------
+
+  async dispatchVerifiedDiscordIngress(
+    context: VerifiedDiscordIngressContext,
+    stream: LegacyRuntimeStream,
+  ): Promise<LegacyRuntimeResult<VerifiedDiscordTurnLease>> {
+    const scope = toScope(context.scope);
+    const { conversationId } = context;
+
+    const resolved = await this.resolveOrCreate(
+      conversationId,
+      scope,
+      { agentConfigId: context.configuredAgent.agentConfigId },
+      {
+        agentConfigId: context.configuredAgent.agentConfigId,
+        instanceId: context.configuredAgent.instanceId,
+      },
+    );
+    if (!resolved.ok) return resolved;
+
+    let detach: () => void;
+    try {
+      detach = this.subscribe(conversationId, scope, stream);
+    } catch (err) {
+      // A partial listener/channel setup rolled itself back inside subscribe(); surface a total
+      // safe failure instead of throwing out of the port. Retryable — the attach is transient.
+      this.logger.error(
+        `Embedded Discord subscription failed for conversation=${conversationId}`,
+        err instanceof Error ? err.message : String(err),
+      );
+      return { ok: false, code: 'runtime_unavailable', retryable: true };
+    }
+
+    return {
+      ok: true,
+      value: this.buildLease(
+        conversationId,
+        scope,
+        context.content,
+        context.attachments,
+        detach,
+        resolved.presentation,
+      ),
+    };
+  }
+
+  // -------------------------------------------------------------------------
+  // Shared helpers
+  // -------------------------------------------------------------------------
+
+  /**
+   * Resolves the owned session, creating it on first use. Ownership/scope rejections
+   * (`Forbidden`/`NotFound`) collapse to `conversation_unavailable`; any other creation
+   * failure surfaces as the retryable `runtime_unavailable`. On success returns the
+   * session presentation so callers avoid a redundant `getSession`.
+   */
+  private async resolveOrCreate(
+    conversationId: string,
+    scope: ActorTenantScope,
+    extraOptions: Readonly<{ provider?: string; modelId?: string; agentConfigId?: string }>,
+    expectedAgent?: Readonly<{ agentConfigId: string; instanceId: string }>,
+  ): Promise<
+    | { readonly ok: true; readonly presentation: LegacySessionPresentation }
+    | Exclude<LegacyRuntimeResult<never>, { ok: true }>
+  > {
+    // A verified-Discord turn may only run under a session whose configured identity matches the
+    // reconciled agent record EXACTLY (config id + resolved name). This holds for BOTH a reused
+    // pre-existing session AND a freshly created one: a session carrying a different configured
+    // agent — however it arose — is rejected rather than executed under the verified label, so we
+    // never silently run a different prompt/model/tool policy. A plain (non-verified) turn passes
+    // no expectedAgent and skips the check.
+    const identityMatches = (candidate: AgentSession): boolean =>
+      expectedAgent === undefined ||
+      (candidate.agentConfigId === expectedAgent.agentConfigId &&
+        candidate.agentName === expectedAgent.instanceId);
+
+    let session = this.agentService.getSession(conversationId, scope);
+    if (session && !identityMatches(session)) {
+      // Reused same-scope session minted under a different configured identity — reject with zero
+      // effects rather than dispatch a verified turn onto a foreign agent's session.
+      return CONVERSATION_UNAVAILABLE;
+    }
+    if (!session) {
+      try {
+        session = await this.agentService.createSession(conversationId, {
+          userId: scope.userId,
+          tenantId: scope.tenantId,
+          ...extraOptions,
+        });
+      } catch (err) {
+        if (err instanceof ForbiddenException || err instanceof NotFoundException) {
+          return CONVERSATION_UNAVAILABLE;
+        }
+        this.logger.error(
+          `Embedded session creation failed for conversation=${conversationId}`,
+          err instanceof Error ? err.stack : String(err),
+        );
+        return { ok: false, code: 'runtime_unavailable', retryable: true };
+      }
+      // The just-created session must ALSO carry the reconciled identity before any effect. A
+      // createSession that returns a session under a different configured agent (misconfiguration
+      // or a substituted factory) is rejected here, before subscribe/persist/ack/prompt.
+      if (!identityMatches(session)) {
+        return CONVERSATION_UNAVAILABLE;
+      }
+    }
+    return { ok: true, presentation: this.presentationForSession(session) };
+  }
+
+  /** Installs a normalizing event listener that forwards to the server-owned stream. */
+  private subscribe(
+    conversationId: string,
+    scope: ActorTenantScope,
+    stream: LegacyRuntimeStream,
+  ): () => void {
+    const unsubscribe = this.agentService.onEvent(
+      conversationId,
+      (event: AgentSessionEvent) => {
+        const normalized = this.normalizeEvent(conversationId, scope, event);
+        if (normalized) stream.onEvent(normalized);
+      },
+      scope,
+    );
+    try {
+      this.agentService.addChannel(conversationId, stream.channelId, scope);
+    } catch (err) {
+      // Partial setup: the listener was acquired but the channel attach failed. Roll back
+      // exactly what was acquired (the listener) before the failure escapes, so no leaked
+      // subscription survives; the caller converts the rethrow into a total safe failure.
+      try {
+        unsubscribe();
+      } catch {
+        /* idempotent teardown */
+      }
+      throw err;
+    }
+    return () => {
+      try {
+        unsubscribe();
+      } catch {
+        /* idempotent teardown */
+      }
+      try {
+        this.agentService.removeChannel(conversationId, stream.channelId, scope);
+      } catch {
+        /* idempotent teardown */
+      }
+    };
+  }
+
+  /** Builds an atomically one-shot, scope-rechecking dispatch lease. */
+  private buildLease(
+    conversationId: string,
+    scope: ActorTenantScope,
+    content: string,
+    attachments: VerifiedDiscordIngressContext['attachments'],
+    detach: () => void,
+    presentation: LegacySessionPresentation,
+  ): LegacySocketTurnLease & VerifiedDiscordTurnLease {
+    let dispatched = false;
+    let disposed = false;
+    return {
+      presentation,
+      dispatch: async (): Promise<LegacyRuntimeResult<void>> => {
+        if (dispatched) {
+          return { ok: false, code: 'turn_already_dispatched', retryable: false };
+        }
+        dispatched = true;
+        try {
+          await this.agentService.prompt(conversationId, content, scope, attachments);
+        } catch (err) {
+          this.logger.error(
+            `Legacy dispatch failed for conversation=${conversationId}`,
+            err instanceof Error ? err.message : String(err),
+          );
+          return { ok: false, code: 'operation_failed', retryable: false };
+        }
+        return { ok: true, value: undefined };
+      },
+      dispose: async (): Promise<void> => {
+        if (disposed) return;
+        disposed = true;
+        detach();
+      },
+    };
+  }
+
+  /** Normalizes a raw agent event into the redaction-agnostic transport event, or drops it. */
+  private normalizeEvent(
+    conversationId: string,
+    scope: ActorTenantScope,
+    event: AgentSessionEvent,
+  ): LegacyRuntimeEvent | undefined {
+    switch (event.type) {
+      case 'agent_start':
+        return { type: 'started' };
+      case 'agent_end':
+        return { type: 'settled', ...this.usageFor(conversationId, scope) };
+      case 'message_update': {
+        const assistant = event.assistantMessageEvent;
+        if (assistant.type === 'text_delta') return { type: 'text_delta', text: assistant.delta };
+        if (assistant.type === 'thinking_delta') {
+          return { type: 'thinking_delta', text: assistant.delta };
+        }
+        return undefined;
+      }
+      case 'tool_execution_start':
+        return { type: 'tool_started', toolCallId: event.toolCallId, toolName: event.toolName };
+      case 'tool_execution_end':
+        return {
+          type: 'tool_finished',
+          toolCallId: event.toolCallId,
+          toolName: event.toolName,
+          isError: event.isError,
+        };
+      default:
+        return undefined;
+    }
+  }
+
+  /**
+   * Gathers terminal usage from the Pi session and records it into session metrics.
+   * Embedded owns AgentService metrics; the gateway never touches `piSession` stats.
+   */
+  private usageFor(conversationId: string, scope: ActorTenantScope): { usage?: LegacyUsage } {
+    const session = this.agentService.getSession(conversationId, scope);
+    const piSession = session?.piSession;
+    const stats = piSession?.getSessionStats();
+    if (!session || !stats) return {};
+    const contextUsage = piSession?.getContextUsage();
+
+    const tokens = {
+      input: stats.tokens?.input ?? 0,
+      output: stats.tokens?.output ?? 0,
+      cacheRead: stats.tokens?.cacheRead ?? 0,
+      cacheWrite: stats.tokens?.cacheWrite ?? 0,
+      total: stats.tokens?.total ?? 0,
+    };
+
+    this.agentService.recordTokenUsage(conversationId, { ...tokens });
+
+    return {
+      usage: {
+        provider: session.provider,
+        modelId: session.modelId,
+        thinkingLevel: piSession?.thinkingLevel ?? 'off',
+        tokens,
+        cost: stats.cost ?? 0,
+        context: {
+          percent: contextUsage?.percent ?? null,
+          window: contextUsage?.contextWindow ?? 0,
+        },
+      },
+    };
+  }
+
+  /** Presentation from a live session id, or undefined when no owned session exists. */
+  private presentationFor(
+    conversationId: string,
+    scope: ActorTenantScope,
+  ): LegacySessionPresentation | undefined {
+    const session = this.agentService.getSession(conversationId, scope);
+    return session ? this.presentationForSession(session) : undefined;
+  }
+
+  /** User-facing projection carrying no session handle, credential, or raw stats. */
+  private presentationForSession(session: AgentSession): LegacySessionPresentation {
+    return {
+      provider: session.provider,
+      modelId: session.modelId,
+      thinkingLevel: session.piSession.thinkingLevel,
+      availableThinkingLevels: session.piSession.getAvailableThinkingLevels(),
+      ...(session.agentName ? { agentName: session.agentName } : {}),
+    };
+  }
+}
+
+/** The shared terminal `conversation_unavailable` failure (missing/foreign/lost ownership). */
+const CONVERSATION_UNAVAILABLE = {
+  ok: false as const,
+  code: 'conversation_unavailable' as const,
+  retryable: false as const,
+};
+
+/** Narrows a branded context scope to the `AgentService` actor/tenant scope (identical shape). */
+function toScope(scope: Readonly<{ userId: string; tenantId: string }>): ActorTenantScope {
+  return { userId: scope.userId, tenantId: scope.tenantId };
+}

apps/gateway/src/chat/harness-chat.runtime.spec.ts

@@ -0,0 +1,170 @@
+import { describe, expect, it } from 'vitest';
+import type {
+  AttachConversation,
+  ConversationSnapshot,
+  DetachConversation,
+  HarnessActorContext,
+  HarnessConversationService,
+  HarnessEventEnvelope,
+  HarnessSelection,
+  SendHarnessTurn,
+  TurnReceipt,
+} from '@mosaicstack/types';
+import { HarnessChatRuntime } from './harness-chat.runtime.js';
+
+/**
+ * Task Five, Step One (harness runtime). Proves the `pi-rpc` runtime executes
+ * exclusively through the {@link HarnessConversationService} RPC boundary and
+ * forwards the caller's exact selection tuple and idempotency key without
+ * substitution. Red-first: the runtime is an unimplemented stub, so every
+ * delegation assertion fails until Step Three.
+ */
+
+const context: HarnessActorContext = {
+  actorId: 'actor-1',
+  tenantId: 'tenant-1',
+  seatId: 'seat-1',
+  correlationId: 'corr-1',
+};
+
+const selection: HarnessSelection = {
+  harnessId: 'pi',
+  providerId: 'anthropic',
+  modelId: 'claude-opus-4-8',
+};
+
+const conversationId = '11111111-1111-4111-8111-111111111111';
+const idempotencyKey = '22222222-2222-4222-8222-222222222222';
+
+const sendInput: SendHarnessTurn & { idempotencyKey: string } = {
+  context,
+  conversationId,
+  selection,
+  turnId: 'turn-abc',
+  correlationId: 'corr-1',
+  content: 'hello',
+  idempotencyKey,
+};
+
+const attachInput: AttachConversation & { afterSequence?: number } = {
+  context,
+  conversationId,
+  clientId: 'client-1',
+  selection,
+  afterSequence: 0,
+};
+
+const detachInput: DetachConversation = {
+  context,
+  conversationId,
+  clientId: 'client-1',
+};
+
+interface RecordedCalls {
+  attach: (AttachConversation & { afterSequence?: number })[];
+  detach: DetachConversation[];
+  send: (SendHarnessTurn & { idempotencyKey: string })[];
+  subscribeFrom: { conversationId: string; afterSequence: number }[];
+}
+
+const snapshot: ConversationSnapshot = {
+  session: {
+    conversationId,
+    nativeSessionId: 'native-1',
+    seatId: 'seat-1',
+    selection,
+    state: 'idle',
+    attachedClientIds: ['client-1'],
+  },
+  lastSequence: 0,
+  replay: [],
+};
+
+function build(): { runtime: HarnessChatRuntime; calls: RecordedCalls } {
+  const calls: RecordedCalls = { attach: [], detach: [], send: [], subscribeFrom: [] };
+  const service: HarnessConversationService = {
+    attach: (input) => {
+      calls.attach.push(input);
+      return Promise.resolve(snapshot);
+    },
+    detach: (input) => {
+      calls.detach.push(input);
+      return Promise.resolve();
+    },
+    send: (input) => {
+      calls.send.push(input);
+      // The service echoes only the requested tuple; there is no representable substitute.
+      const receipt: TurnReceipt = {
+        conversationId: input.conversationId,
+        turnId: 'turn-server',
+        correlationId: input.correlationId,
+        state: 'accepted',
+        selection: input.selection,
+      };
+      return Promise.resolve(receipt);
+    },
+    subscribeFrom: (id, afterSequence) => {
+      calls.subscribeFrom.push({ conversationId: id, afterSequence });
+
+      return (async function* (): AsyncIterable<HarnessEventEnvelope> {
+        return;
+      })();
+    },
+  };
+  return { runtime: new HarnessChatRuntime(service), calls };
+}
+
+describe('HarnessChatRuntime', () => {
+  it('is the harness runtime kind and needs only a HarnessConversationService', () => {
+    const { runtime } = build();
+    expect(runtime.kind).toBe('harness');
+  });
+
+  it('delegates send to the conversation service with the exact tuple and idempotency key', async () => {
+    const { runtime, calls } = build();
+
+    const receipt = await runtime.send(sendInput);
+
+    expect(calls.send).toHaveLength(1);
+    const firstSend = calls.send[0]!;
+    expect(firstSend).toEqual(sendInput);
+    expect(firstSend.idempotencyKey).toBe(idempotencyKey);
+    expect(firstSend.selection).toEqual(selection);
+    // The runtime must not substitute an effective tuple onto the receipt.
+    expect(receipt.selection).toEqual(selection);
+  });
+
+  it('delegates attach to the conversation service and returns its snapshot', async () => {
+    const { runtime, calls } = build();
+
+    const result = await runtime.attach(attachInput);
+
+    expect(calls.attach).toHaveLength(1);
+    expect(calls.attach[0]).toEqual(attachInput);
+    expect(result).toBe(snapshot);
+  });
+
+  it('delegates detach to the conversation service', async () => {
+    const { runtime, calls } = build();
+
+    await runtime.detach(detachInput);
+
+    expect(calls.detach).toHaveLength(1);
+    expect(calls.detach[0]).toEqual(detachInput);
+  });
+
+  it('delegates subscribeFrom to the conversation service journal replay', async () => {
+    const { runtime, calls } = build();
+
+    const iterable = runtime.subscribeFrom(conversationId, 7);
+    // Drain to prove it is the service-backed async iterable, not a fabricated one.
+    const drained: unknown[] = [];
+    for await (const event of iterable) {
+      drained.push(event);
+    }
+    expect(drained).toHaveLength(0);
+
+    expect(calls.subscribeFrom).toHaveLength(1);
+    expect(calls.subscribeFrom[0]).toEqual({ conversationId, afterSequence: 7 });
+  });
+});

apps/gateway/src/chat/harness-chat.runtime.ts

@@ -0,0 +1,47 @@
+import type {
+  AttachConversation,
+  ConversationSnapshot,
+  DetachConversation,
+  HarnessConversationService,
+  HarnessEventEnvelope,
+  SendHarnessTurn,
+  TurnReceipt,
+} from '@mosaicstack/types';
+import type { ChatRuntime } from './chat-runtime.js';
+
+/**
+ * The `pi-rpc` chat runtime. It executes browser chat exclusively through the
+ * harness-neutral {@link HarnessConversationService} RPC boundary — it never
+ * touches the embedded `AgentService`/`ProviderService`/`RoutingEngineService`
+ * stack, and it forwards the caller's exact selection tuple and idempotency key
+ * without substitution.
+ *
+ * It owns no state and adds no policy: every method forwards the caller's exact
+ * argument to the injected {@link HarnessConversationService} and returns its
+ * result unchanged, so the requested selection tuple and idempotency key can
+ * never be substituted on the way through.
+ */
+export class HarnessChatRuntime implements ChatRuntime {
+  readonly kind = 'harness' as const;
+
+  constructor(private readonly conversations: HarnessConversationService) {}
+
+  attach(input: AttachConversation & { afterSequence?: number }): Promise<ConversationSnapshot> {
+    return this.conversations.attach(input);
+  }
+
+  detach(input: DetachConversation): Promise<void> {
+    return this.conversations.detach(input);
+  }
+
+  send(input: SendHarnessTurn & { idempotencyKey: string }): Promise<TurnReceipt> {
+    return this.conversations.send(input);
+  }
+
+  subscribeFrom(
+    conversationId: string,
+    afterSequence: number,
+  ): AsyncIterable<HarnessEventEnvelope> {
+    return this.conversations.subscribeFrom(conversationId, afterSequence);
+  }
+}

apps/gateway/src/commands/command-executor-p8012.spec.ts

@@ -1,3 +1,4 @@
+import { Logger } from '@nestjs/common';
 import { describe, it, expect, vi, beforeEach } from 'vitest';
 import { CommandExecutorService } from './command-executor.service.js';
 import type { SlashCommandPayload } from '@mosaicstack/types';
@@ -12,6 +13,7 @@ const mockRegistry = {
       { name: 'agent', aliases: ['a'], scope: 'agent', execution: 'socket', available: true },
       { name: 'prdy', aliases: [], scope: 'agent', execution: 'socket', available: true },
       { name: 'tools', aliases: [], scope: 'agent', execution: 'socket', available: true },
+      { name: 'mcp', aliases: [], scope: 'agent', execution: 'socket', available: true },
     ],
     skills: [],
   })),
@@ -72,7 +74,20 @@ const mockChatGateway = {
   broadcastSessionInfo: vi.fn(),
 };
 
-function buildService(redis: typeof mockRedis | null = mockRedis): CommandExecutorService {
+const mockMcpClient = {
+  reconnectServer: vi.fn().mockResolvedValue(undefined),
+  getServerStatuses: vi.fn(() => []),
+  getToolDefinitions: vi.fn(() => []),
+};
+
+function buildService(
+  redis: typeof mockRedis | null = mockRedis,
+  mcpClient: {
+    reconnectServer: ReturnType<typeof vi.fn>;
+    getServerStatuses: ReturnType<typeof vi.fn>;
+    getToolDefinitions: ReturnType<typeof vi.fn>;
+  } = mockMcpClient,
+): CommandExecutorService {
   return new CommandExecutorService(
     mockRegistry as never,
     mockAgentService as never,
@@ -82,7 +97,7 @@ function buildService(redis: typeof mockRedis | null = mockRedis): CommandExecut
     mockBrain as never,
     null,
     mockChatGateway as never,
-    null,
+    mcpClient as never,
   );
 }
 
@@ -258,4 +273,124 @@ describe('CommandExecutorService — P8-012 commands', () => {
     expect(result.command).toBe('tools');
     expect(result.message).toContain('tools');
   });
+
+  // Top-level catch sanitization (P3-4 re-review finding #1): a rejected
+  // Redis `set` inside /provider login is the only reachable path into the
+  // top-level catch in `execute()`. The raw exception must be logged
+  // server-side but never handed back to the socket client.
+  it('sanitizes the top-level command catch, logging the raw exception but never returning it to the client', async () => {
+    const distinctiveRawFailure = 'ECONNREFUSED distinctive-raw-redis-failure-token-9f31';
+    const rawError = new Error(distinctiveRawFailure);
+    const failingRedis = {
+      set: vi.fn().mockRejectedValue(rawError),
+      get: vi.fn(),
+      del: vi.fn(),
+    };
+    const failingService = buildService(failingRedis as unknown as typeof mockRedis);
+    const loggerErrorSpy = vi.spyOn(Logger.prototype, 'error').mockImplementation(() => undefined);
+
+    const payload: SlashCommandPayload = {
+      command: 'provider',
+      args: 'login anthropic',
+      conversationId,
+    };
+    const result = await failingService.execute(payload, userScope);
+
+    expect(result.success).toBe(false);
+    expect(result.command).toBe('provider');
+    expect(result.message).toBe('Command failed due to an internal error.');
+    expect(result.message).not.toContain(distinctiveRawFailure);
+    expect(result.message).not.toContain('ECONNREFUSED');
+
+    // The real exception is still logged server-side, as the raw Error
+    // object itself (not stringified/interpolated into the log message).
+    expect(loggerErrorSpy).toHaveBeenCalled();
+    const loggedRawError = loggerErrorSpy.mock.calls.some((call) => call.includes(rawError));
+    expect(loggedRawError).toBe(true);
+
+    loggerErrorSpy.mockRestore();
+  });
+
+  // Inner catch sanitization (P3-5 operator ruling): every catch in
+  // command-executor.service.ts that returns a SlashCommandResultPayload
+  // must sanitize the client-facing message the same way the top-level
+  // catch does, while still logging the raw exception server-side.
+  it('/agent new sanitizes agent-creation failures, logging the raw exception but never returning it to the client', async () => {
+    const marker = new Error('distinctive-agent-create-failure-token-A17f');
+    mockBrain.agents.create.mockRejectedValueOnce(marker);
+    const loggerErrorSpy = vi.spyOn(Logger.prototype, 'error').mockImplementation(() => undefined);
+
+    const payload: SlashCommandPayload = {
+      command: 'agent',
+      args: 'new my-new-agent',
+      conversationId,
+    };
+    const result = await service.execute(payload, userScope);
+
+    expect(result.success).toBe(false);
+    expect(result.command).toBe('agent');
+    expect(result.message).toBe('Failed to create agent due to an internal error.');
+    expect(result.message).not.toContain('distinctive-agent-create-failure-token-A17f');
+
+    expect(loggerErrorSpy).toHaveBeenCalled();
+    const loggedRawError = loggerErrorSpy.mock.calls.some((call) => call.includes(marker));
+    expect(loggedRawError).toBe(true);
+
+    loggerErrorSpy.mockRestore();
+  });
+
+  it('/agent <name> switch sanitizes agent-lookup failures, logging the raw exception but never returning it to the client', async () => {
+    const marker = new Error('distinctive-agent-switch-failure-token-B29c');
+    mockBrain.agents.findByName.mockRejectedValueOnce(marker);
+    const loggerErrorSpy = vi.spyOn(Logger.prototype, 'error').mockImplementation(() => undefined);
+
+    const payload: SlashCommandPayload = {
+      command: 'agent',
+      args: 'some-other-agent',
+      conversationId,
+    };
+    const result = await service.execute(payload, userScope);
+
+    expect(result.success).toBe(false);
+    expect(result.command).toBe('agent');
+    expect(result.message).toBe('Failed to switch agent due to an internal error.');
+    expect(result.message).not.toContain('distinctive-agent-switch-failure-token-B29c');
+
+    expect(loggerErrorSpy).toHaveBeenCalled();
+    const loggedRawError = loggerErrorSpy.mock.calls.some((call) => call.includes(marker));
+    expect(loggedRawError).toBe(true);
+
+    loggerErrorSpy.mockRestore();
+  });
+
+  it('/mcp reconnect sanitizes MCP client failures, logging the raw exception but never returning it to the client', async () => {
+    const marker = new Error('distinctive-mcp-reconnect-failure-token-C33e');
+    const mockMcpClient = {
+      reconnectServer: vi.fn().mockRejectedValue(marker),
+      getServerStatuses: vi.fn(() => []),
+      getToolDefinitions: vi.fn(() => []),
+    };
+    const mcpService = buildService(mockRedis, mockMcpClient);
+    const loggerErrorSpy = vi.spyOn(Logger.prototype, 'error').mockImplementation(() => undefined);
+
+    const payload: SlashCommandPayload = {
+      command: 'mcp',
+      args: 'reconnect my-server',
+      conversationId,
+    };
+    const result = await mcpService.execute(payload, userScope);
+
+    expect(result.success).toBe(false);
+    expect(result.command).toBe('mcp');
+    expect(result.message).toBe(
+      'Failed to reconnect MCP server "my-server" due to an internal error.',
+    );
+    expect(result.message).not.toContain('distinctive-mcp-reconnect-failure-token-C33e');
+
+    expect(loggerErrorSpy).toHaveBeenCalled();
+    const loggedRawError = loggerErrorSpy.mock.calls.some((call) => call.includes(marker));
+    expect(loggedRawError).toBe(true);
+
+    loggerErrorSpy.mockRestore();
+  });
 });

apps/gateway/src/commands/command-executor-tess-security.spec.ts

@@ -36,6 +36,12 @@ const authorization = {
   ),
 };
 
+const mockMcpClient = {
+  getServerStatuses: vi.fn(() => []),
+  getToolDefinitions: vi.fn(() => []),
+  reconnectServer: vi.fn().mockResolvedValue(undefined),
+};
+
 function buildExecutor(authorizationService: unknown = authorization): CommandExecutorService {
   return new CommandExecutorService(
     registry as never,
@@ -46,7 +52,7 @@ function buildExecutor(authorizationService: unknown = authorization): CommandEx
     { agents: {} } as never,
     null,
     null,
-    null,
+    mockMcpClient as never,
     authorizationService as never,
   );
 }

apps/gateway/src/commands/command-executor.service.ts

@@ -34,9 +34,7 @@ export class CommandExecutorService {
     @Optional()
     @Inject(forwardRef(() => ChatGateway))
     private readonly chatGateway: ChatGateway | null,
-    @Optional()
+    @Inject(McpClientService) private readonly mcpClient: McpClientService,
-    @Inject(McpClientService)
-    private readonly mcpClient: McpClientService | null,
     @Optional()
     @Inject(CommandAuthorizationService)
     private readonly authorization: CommandAuthorizationService | null = null,
@@ -159,8 +157,13 @@ export class CommandExecutorService {
           };
       }
     } catch (err) {
-      this.logger.error(`Command /${command} failed: ${err}`);
+      this.logger.error(`Command /${command} failed`, err);
-      return { command, conversationId, success: false, message: String(err) };
+      return {
+        command,
+        conversationId,
+        success: false,
+        message: 'Command failed due to an internal error.',
+      };
     }
   }
 
@@ -336,11 +339,11 @@ export class CommandExecutorService {
           data: { agentId: newAgent.id, agentName: newAgent.name },
         };
       } catch (err) {
-        this.logger.error(`Failed to create agent: ${err}`);
+        this.logger.error(`Failed to create agent "${namePart}" for user ${userId}`, err);
         return {
           command: 'agent',
           success: false,
-          message: `Failed to create agent: ${String(err)}`,
+          message: 'Failed to create agent due to an internal error.',
           conversationId,
         };
       }
@@ -391,11 +394,11 @@ export class CommandExecutorService {
         data: { agentId: agentConfig.id, agentName: agentConfig.name, model: agentConfig.model },
       };
     } catch (err) {
-      this.logger.error(`Failed to switch agent "${agentName}": ${err}`);
+      this.logger.error(`Failed to switch agent "${agentName}"`, err);
       return {
         command: 'agent',
         success: false,
-        message: `Failed to switch agent: ${String(err)}`,
+        message: 'Failed to switch agent due to an internal error.',
         conversationId,
       };
     }
@@ -543,15 +546,6 @@ export class CommandExecutorService {
     args: string | null,
     conversationId: string,
   ): Promise<SlashCommandResultPayload> {
-    if (!this.mcpClient) {
-      return {
-        command: 'mcp',
-        conversationId,
-        success: false,
-        message: 'MCP client service is not available.',
-      };
-    }
-
     const action = args?.trim().split(/\s+/)[0] ?? 'status';
 
     switch (action) {
@@ -608,11 +602,12 @@ export class CommandExecutorService {
             message: `MCP server "${serverName}" reconnected successfully.`,
           };
         } catch (err) {
+          this.logger.error(`Failed to reconnect MCP server "${serverName}"`, err);
           return {
             command: 'mcp',
             conversationId,
             success: false,
-            message: `Failed to reconnect MCP server "${serverName}": ${err instanceof Error ? err.message : String(err)}`,
+            message: `Failed to reconnect MCP server "${serverName}" due to an internal error.`,
           };
         }
       }

apps/gateway/src/commands/commands.integration.spec.ts

@@ -11,6 +11,8 @@
  * - Unknown command returns descriptive error
  */
 import { describe, it, expect, vi, beforeEach } from 'vitest';
+import { CommandsModule } from './commands.module.js';
+import { McpClientModule } from '../mcp-client/mcp-client.module.js';
 import { CommandRegistryService } from './command-registry.service.js';
 import { CommandExecutorService } from './command-executor.service.js';
 import type { SlashCommandPayload } from '@mosaicstack/types';
@@ -47,6 +49,12 @@ const mockBrain = {
   },
 };
 
+const mockMcpClient = {
+  getServerStatuses: vi.fn(() => []),
+  getToolDefinitions: vi.fn(() => []),
+  reconnectServer: vi.fn().mockResolvedValue(undefined),
+};
+
 // ─── Helpers ─────────────────────────────────────────────────────────────────
 
 function buildRegistry(): CommandRegistryService {
@@ -65,7 +73,7 @@ function buildExecutor(registry: CommandRegistryService): CommandExecutorService
     mockBrain as never,
     null, // reloadService (optional)
     null, // chatGateway (optional)
-    null, // mcpClient (optional)
+    mockMcpClient as never,
   );
 }
 
@@ -153,6 +161,15 @@ describe('CommandRegistryService — integration', () => {
   }
 });
 
+// ─── Module Wiring Tests ──────────────────────────────────────────────────────
+
+describe('CommandsModule — Nest wiring', () => {
+  it('CommandsModule imports McpClientModule in its Nest metadata', () => {
+    const imports = Reflect.getMetadata('imports', CommandsModule) ?? [];
+    expect(imports).toContain(McpClientModule);
+  });
+});
+
 // ─── Executor Tests ───────────────────────────────────────────────────────────
 
 describe('CommandExecutorService — integration', () => {
@@ -259,4 +276,14 @@ describe('CommandExecutorService — integration', () => {
       expect(result.command).toBe(cmd);
     });
   }
+
+  // /mcp status reaches the required McpClientService and never reports it unavailable
+  it('/mcp status calls the wired McpClientService and reports the no-servers message', async () => {
+    const payload: SlashCommandPayload = { command: 'mcp', conversationId };
+    const result = await executor.execute(payload, userScope);
+    expect(mockMcpClient.getServerStatuses).toHaveBeenCalledOnce();
+    expect(result.success).toBe(true);
+    expect(result.message).toContain('No MCP servers configured.');
+    expect(result.message).not.toBe('MCP client service is not available.');
+  });
 });

apps/gateway/src/commands/commands.module.ts

@@ -4,6 +4,7 @@ import type { MosaicConfig } from '@mosaicstack/config';
 import { MOSAIC_CONFIG } from '../config/config.module.js';
 import { ChatModule } from '../chat/chat.module.js';
 import { GCModule } from '../gc/gc.module.js';
+import { McpClientModule } from '../mcp-client/mcp-client.module.js';
 import { ReloadModule } from '../reload/reload.module.js';
 import { CommandAuthorizationService } from './command-authorization.service.js';
 import { CommandExecutorService } from './command-executor.service.js';
@@ -14,7 +15,12 @@ import { COMMANDS_REDIS } from './commands.tokens.js';
 const COMMANDS_QUEUE_HANDLE = 'COMMANDS_QUEUE_HANDLE';
 
 @Module({
-  imports: [GCModule, forwardRef(() => ReloadModule), forwardRef(() => ChatModule)],
+  imports: [
+    GCModule,
+    McpClientModule,
+    forwardRef(() => ReloadModule),
+    forwardRef(() => ChatModule),
+  ],
   providers: [
     {
       provide: COMMANDS_QUEUE_HANDLE,

apps/gateway/src/config/config.module.ts

@@ -1,5 +1,6 @@
 import { Global, Module } from '@nestjs/common';
 import { loadConfig, type MosaicConfig } from '@mosaicstack/config';
+import { resolveGatewayConfigPath } from '../env.js';
 
 export const MOSAIC_CONFIG = 'MOSAIC_CONFIG';
 
@@ -8,7 +9,7 @@ export const MOSAIC_CONFIG = 'MOSAIC_CONFIG';
   providers: [
     {
       provide: MOSAIC_CONFIG,
-      useFactory: (): MosaicConfig => loadConfig(),
+      useFactory: (): MosaicConfig => loadConfig(resolveGatewayConfigPath()),
     },
   ],
   exports: [MOSAIC_CONFIG],

apps/gateway/src/conversations/conversations-harness-fence.spec.ts

@@ -0,0 +1,116 @@
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+import type { ChatRuntimeMode } from '../chat/chat-runtime.js';
+import { ConversationsController } from './conversations.controller.js';
+
+/**
+ * Task 5 harness fence for the conversations REST write path.
+ *
+ * Under `pi-rpc` the durable/harness conversation path (Task 15) owns message persistence, so the
+ * legacy direct-repository write via `POST /api/conversations/:id/messages` must be refused with a
+ * fixed typed `runtime_unsupported` BEFORE the repository is touched — never a duplicate write.
+ * Under `legacy` the endpoint keeps its current behaviour and writes through `brain.conversations`.
+ *
+ * Item 3 (single runtime-mode source of truth): the mode is the router's ONE init-time resolution,
+ * injected into the controller and read as `router.runtimeMode`. It is NOT re-derived from
+ * `process.env` at request time. The two "env is flipped after construction" tests below are the
+ * load-bearing guard: they pass only because the controller reads the fixed injected mode, and turn
+ * RED the instant the fence is reverted to `resolveChatRuntimeMode(process.env)`.
+ */
+const CONVERSATION_ID = '22222222-2222-4222-8222-222222222222';
+const USER = { id: 'user-1' };
+
+function sendMessageDto() {
+  return {
+    role: 'user' as const,
+    content: 'hello from the legacy REST write path',
+    metadata: undefined,
+  };
+}
+
+function brainWithMessageSpy() {
+  const addMessage = vi.fn().mockResolvedValue({
+    id: 'message-1',
+    conversationId: CONVERSATION_ID,
+    role: 'user',
+    content: 'hello from the legacy REST write path',
+  });
+  return {
+    brain: { conversations: { addMessage } } as never,
+    addMessage,
+  };
+}
+
+/** The controller only needs the router's immutable `runtimeMode`; supply exactly that. */
+function routerFixedTo(mode: ChatRuntimeMode) {
+  return { runtimeMode: mode };
+}
+
+let priorMode: string | undefined;
+
+describe('conversations REST write path — Task 5 harness fence', () => {
+  beforeEach(() => {
+    priorMode = process.env['CHAT_HARNESS_RUNTIME'];
+  });
+
+  afterEach(() => {
+    if (priorMode === undefined) delete process.env['CHAT_HARNESS_RUNTIME'];
+    else process.env['CHAT_HARNESS_RUNTIME'] = priorMode;
+  });
+
+  it('refuses the legacy repository write when the router resolved pi-rpc, before any write', async () => {
+    const { brain, addMessage } = brainWithMessageSpy();
+    const controller = new ConversationsController(brain, routerFixedTo('pi-rpc'));
+
+    await expect(
+      controller.addMessage(CONVERSATION_ID, sendMessageDto(), USER),
+    ).rejects.toMatchObject({ code: 'runtime_unsupported' });
+
+    // Load-bearing: the durable/harness path owns pi-rpc persistence — the legacy repo must not be
+    // written, so no duplicate message can be produced.
+    expect(addMessage).not.toHaveBeenCalled();
+  });
+
+  it('writes through the repository when the router resolved legacy (GREEN control)', async () => {
+    const { brain, addMessage } = brainWithMessageSpy();
+    const controller = new ConversationsController(brain, routerFixedTo('legacy'));
+
+    const result = await controller.addMessage(CONVERSATION_ID, sendMessageDto(), USER);
+
+    expect(addMessage).toHaveBeenCalledWith(
+      {
+        conversationId: CONVERSATION_ID,
+        role: 'user',
+        content: 'hello from the legacy REST write path',
+        metadata: undefined,
+      },
+      USER.id,
+    );
+    expect(result).toMatchObject({ id: 'message-1', conversationId: CONVERSATION_ID });
+  });
+
+  it('keeps refusing under a pi-rpc router even when CHAT_HARNESS_RUNTIME is flipped to legacy after startup', async () => {
+    // The runtime mode is fixed at module init. A later env mutation must not reopen the fence:
+    // a request-time `resolveChatRuntimeMode(process.env)` read would see `legacy` and wrongly write.
+    process.env['CHAT_HARNESS_RUNTIME'] = 'legacy';
+    const { brain, addMessage } = brainWithMessageSpy();
+    const controller = new ConversationsController(brain, routerFixedTo('pi-rpc'));
+
+    await expect(
+      controller.addMessage(CONVERSATION_ID, sendMessageDto(), USER),
+    ).rejects.toMatchObject({ code: 'runtime_unsupported' });
+
+    expect(addMessage).not.toHaveBeenCalled();
+  });
+
+  it('keeps writing under a legacy router even when CHAT_HARNESS_RUNTIME is flipped to pi-rpc after startup', async () => {
+    // Symmetric guard: a legacy-resolved router must keep writing regardless of the live env, so a
+    // request-time env read of `pi-rpc` cannot spuriously refuse a legitimate legacy write.
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
+    const { brain, addMessage } = brainWithMessageSpy();
+    const controller = new ConversationsController(brain, routerFixedTo('legacy'));
+
+    await controller.addMessage(CONVERSATION_ID, sendMessageDto(), USER);
+
+    expect(addMessage).toHaveBeenCalledTimes(1);
+  });
+});

apps/gateway/src/conversations/conversations.controller.ts

@@ -6,6 +6,7 @@ import {
   ForbiddenException,
   Get,
   HttpCode,
+  HttpException,
   HttpStatus,
   Inject,
   NotFoundException,
@@ -19,6 +20,7 @@ import type { Brain } from '@mosaicstack/brain';
 import { BRAIN } from '../brain/brain.tokens.js';
 import { AuthGuard } from '../auth/auth.guard.js';
 import { CurrentUser } from '../auth/current-user.decorator.js';
+import { ChatRuntimeRouter } from '../chat/chat-runtime-router.js';
 import {
   CreateConversationDto,
   UpdateConversationDto,
@@ -26,10 +28,41 @@ import {
   SearchMessagesDto,
 } from './conversations.dto.js';
 
+/**
+ * Under `pi-rpc` the durable/harness conversation path (Task 15) owns message persistence, so the
+ * legacy direct-repository write must fail closed with a fixed typed `runtime_unsupported` before
+ * the repository is touched — never a duplicate write. The `code` field is exposed at the top level
+ * so callers can discriminate the refusal while the 503 status carries the browser-safe surface.
+ */
+class HarnessRuntimeWriteUnsupportedException extends HttpException {
+  readonly code = 'runtime_unsupported' as const;
+
+  constructor() {
+    super(
+      {
+        code: 'runtime_unsupported',
+        message:
+          'Conversation message writes are handled by the harness runtime on this deployment.',
+      },
+      HttpStatus.SERVICE_UNAVAILABLE,
+    );
+  }
+}
+
 @Controller('api/conversations')
 @UseGuards(AuthGuard)
 export class ConversationsController {
-  constructor(@Inject(BRAIN) private readonly brain: Brain) {}
+  /**
+   * `router` supplies the ONE immutable runtime mode resolved at module init (Task 5, item 3).
+   * The pre-write fence reads `router.runtimeMode`, never `resolveChatRuntimeMode(process.env)` at
+   * request time — a single source of truth, so the controller cannot disagree with the router
+   * about the live runtime if the environment is mutated after startup. Narrowed to `runtimeMode`
+   * so this class depends on nothing else the router exposes.
+   */
+  constructor(
+    @Inject(BRAIN) private readonly brain: Brain,
+    @Inject(ChatRuntimeRouter) private readonly router: Pick<ChatRuntimeRouter, 'runtimeMode'>,
+  ) {}
 
   @Get()
   async list(@CurrentUser() user: { id: string }) {
@@ -94,6 +127,13 @@ export class ConversationsController {
     @Body() dto: SendMessageDto,
     @CurrentUser() user: { id: string },
   ) {
+    // Fail the legacy repository write closed under pi-rpc BEFORE touching the repository — the
+    // harness path owns persistence there, so a direct write would duplicate the message. The mode
+    // comes from the router's init-time resolution, not a request-time env read.
+    if (this.router.runtimeMode === 'pi-rpc') {
+      throw new HarnessRuntimeWriteUnsupportedException();
+    }
+
     const message = await this.brain.conversations.addMessage(
       {
         conversationId: id,

apps/gateway/src/conversations/conversations.module.ts

@@ -1,7 +1,14 @@
 import { Module } from '@nestjs/common';
+import { ChatModule } from '../chat/chat.module.js';
 import { ConversationsController } from './conversations.controller.js';
 
+/**
+ * Imports {@link ChatModule} solely to inject its exported {@link ChatRuntimeRouter} into
+ * {@link ConversationsController}, so the REST write fence reads the same init-time runtime mode the
+ * router resolved — one source of truth, no duplicate provider, no global token, no AppModule edit.
+ */
 @Module({
+  imports: [ChatModule],
   controllers: [ConversationsController],
 })
 export class ConversationsModule {}

apps/gateway/src/coord/interaction-coordination.di.test.ts

@@ -0,0 +1,19 @@
+import 'reflect-metadata';
+import { Test } from '@nestjs/testing';
+import { describe, expect, it } from 'vitest';
+import { CoordModule } from './coord.module.js';
+import { InteractionCoordinationService } from './interaction-coordination.service.js';
+import { AuthGuard } from '../auth/auth.guard.js';
+
+describe('CoordModule DI (compiled-metadata boot)', () => {
+  it('resolves InteractionCoordinationService through Nest DI', async () => {
+    const moduleRef = await Test.createTestingModule({ imports: [CoordModule] })
+      .overrideGuard(AuthGuard)
+      .useValue({ canActivate: (): boolean => true })
+      .compile();
+    expect(moduleRef.get(InteractionCoordinationService)).toBeInstanceOf(
+      InteractionCoordinationService,
+    );
+    await moduleRef.close();
+  });
+});

apps/gateway/src/coord/interaction-coordination.service.ts

@@ -1,4 +1,4 @@
-import { Inject, Injectable } from '@nestjs/common';
+import { Inject, Injectable, Optional } from '@nestjs/common';
 import {
   InteractionCoordinationClient,
   type CoordinationObservation,
@@ -13,6 +13,7 @@ import type { CreateHandoffDto } from './interaction-coordination.dto.js';
 
 export const COORDINATION_PORT = Symbol('COORDINATION_PORT');
 export const COORDINATION_CONFIG = Symbol('COORDINATION_CONFIG');
+export const HANDOFF_ID_FACTORY = Symbol('HANDOFF_ID_FACTORY');
 
 const HANDOFF_TRACKING_TTL_MS = 60 * 60 * 1_000;
 const MAX_TRACKED_HANDOFFS = 1_000;
@@ -60,6 +61,8 @@ export class InteractionCoordinationService {
   constructor(
     @Inject(COORDINATION_PORT) private readonly port: InteractionCoordinationPort,
     @Inject(COORDINATION_CONFIG) private readonly config: InteractionCoordinationConfig,
+    @Optional()
+    @Inject(HANDOFF_ID_FACTORY)
     private readonly handoffIdFactory: () => string = (): string => crypto.randomUUID(),
   ) {}
 

apps/gateway/src/env.ts

@@ -0,0 +1,133 @@
+import { config } from 'dotenv';
+import { existsSync } from 'node:fs';
+import { homedir } from 'node:os';
+import { dirname, join, resolve } from 'node:path';
+import { fileURLToPath } from 'node:url';
+import { detectFromEnv, loadConfig } from '@mosaicstack/config';
+
+type TierSource =
+  | 'process environment'
+  | 'daemon .env'
+  | 'monorepo-root .env'
+  | 'gateway-local .env'
+  | 'default';
+
+type BootSource = TierSource | 'mosaic.config.json';
+
+export interface GatewayDotenvPaths {
+  daemonEnv: string;
+  monorepoRootEnv: string;
+  gatewayLocalEnv: string;
+}
+
+const here = dirname(fileURLToPath(import.meta.url));
+
+export function resolveGatewayDotenvPaths(
+  anchor: string = here,
+  homeBase: string = homedir(),
+): GatewayDotenvPaths {
+  return {
+    daemonEnv: join(homeBase, '.config', 'mosaic', 'gateway', '.env'),
+    monorepoRootEnv: resolve(anchor, '../../..', '.env'),
+    gatewayLocalEnv: resolve(anchor, '..', '.env'),
+  };
+}
+
+export function resolveGatewayConfigPath(anchor: string = here): string {
+  // GATEWAY_HOME is daemon-created 0700; its env override adds no authority because env can set MOSAIC_STORAGE_TIER.
+  const gatewayHome = resolve(
+    process.env['MOSAIC_GATEWAY_HOME'] ?? join(homedir(), '.config', 'mosaic', 'gateway'),
+  );
+  const daemonConfig = join(gatewayHome, 'mosaic.config.json');
+  const gatewayLocalConfig = resolve(anchor, '..', 'mosaic.config.json');
+  const monorepoRootConfig = resolve(anchor, '../../..', 'mosaic.config.json');
+
+  if (existsSync(daemonConfig)) {
+    return daemonConfig;
+  }
+  if (existsSync(gatewayLocalConfig)) {
+    return gatewayLocalConfig;
+  }
+  if (existsSync(monorepoRootConfig)) {
+    return monorepoRootConfig;
+  }
+
+  return monorepoRootConfig;
+}
+
+export function loadGatewayEnv(anchor: string = here, homeBase: string = homedir()): void {
+  const { daemonEnv, monorepoRootEnv, gatewayLocalEnv } = resolveGatewayDotenvPaths(
+    anchor,
+    homeBase,
+  );
+  const inheritedTier = process.env['MOSAIC_STORAGE_TIER'];
+  let tierSource: TierSource = inheritedTier === undefined ? 'default' : 'process environment';
+  const inheritedDatabaseUrl = process.env['DATABASE_URL'];
+  let databaseUrlSource: TierSource =
+    inheritedDatabaseUrl === undefined ? 'default' : 'process environment';
+
+  function loadAnchoredDotenv(
+    path: string,
+    sourceLabel: Exclude<TierSource, 'process environment' | 'default'>,
+  ): void {
+    if (!existsSync(path)) {
+      return;
+    }
+
+    const beforeTier = process.env['MOSAIC_STORAGE_TIER'];
+    const beforeDatabaseUrl = process.env['DATABASE_URL'];
+    config({ path, quiet: true });
+
+    if (
+      beforeTier === undefined &&
+      process.env['MOSAIC_STORAGE_TIER'] !== undefined &&
+      tierSource === 'default'
+    ) {
+      tierSource = sourceLabel;
+    }
+
+    if (
+      beforeDatabaseUrl === undefined &&
+      process.env['DATABASE_URL'] !== undefined &&
+      databaseUrlSource === 'default'
+    ) {
+      databaseUrlSource = sourceLabel;
+    }
+  }
+
+  // Load .env from daemon config dir (global install / daemon mode) first.
+  // It takes precedence over file-based local-dev configuration.
+  loadAnchoredDotenv(daemonEnv, 'daemon .env');
+
+  // Load .env from the anchored monorepo root, then fill any remaining values
+  // from apps/gateway/.env when present.
+  loadAnchoredDotenv(monorepoRootEnv, 'monorepo-root .env');
+  loadAnchoredDotenv(gatewayLocalEnv, 'gateway-local .env');
+
+  const envOnlyTier = detectFromEnv().tier;
+  const configPath = resolveGatewayConfigPath(anchor);
+  const anchoredConfigExists = existsSync(configPath);
+  const resolvedTier = loadConfig(configPath).tier;
+  const configuredTier = process.env['MOSAIC_STORAGE_TIER'];
+  const databaseUrlDeterminesTier = envOnlyTier === 'standalone' && configuredTier !== 'standalone';
+  const recognizedTierDeterminesTier =
+    (configuredTier === 'federated' ||
+      configuredTier === 'standalone' ||
+      configuredTier === 'local') &&
+    configuredTier === envOnlyTier;
+
+  let source: BootSource;
+  if (anchoredConfigExists) {
+    source = 'mosaic.config.json';
+  } else if (databaseUrlDeterminesTier && databaseUrlSource !== 'default') {
+    source = databaseUrlSource;
+  } else if (recognizedTierDeterminesTier && tierSource !== 'default') {
+    source = tierSource;
+  } else {
+    source = 'default';
+  }
+
+  console.info(`[gateway env] storage tier=${resolvedTier} source=${source}`);
+}
+
+loadGatewayEnv();

apps/gateway/src/harness/harness-selection.controller.spec.ts

@@ -0,0 +1,164 @@
+import 'reflect-metadata';
+import {
+  type CanActivate,
+  type ExecutionContext,
+  type INestApplication,
+  ValidationPipe,
+} from '@nestjs/common';
+import { FastifyAdapter, type NestFastifyApplication } from '@nestjs/platform-fastify';
+import { Test } from '@nestjs/testing';
+import request from 'supertest';
+import { afterAll, beforeAll, beforeEach, describe, expect, it } from 'vitest';
+import { AuthGuard } from '../auth/auth.guard.js';
+import { HarnessRegistry } from './harness.registry.js';
+import { HARNESS_REGISTRY } from './harness.tokens.js';
+import { HarnessSelectionRepository } from './harness-selection.repository.js';
+import { FakeHarnessAdapter } from './testing/fake-harness.adapter.js';
+// Import the REAL module (not a hand-listed controllers+mocks list) so an
+// unresolved provider fails at app.init() — the #1145-class DI-boot guard.
+import { HarnessModule } from './harness.module.js';
+
+// A known-available tuple from the fake adapter's default catalog.
+const VALID = { harnessId: 'fake', providerId: 'fake-openai', modelId: 'fake-mini' };
+// A tuple whose provider/model are not in any catalog.
+const UNKNOWN = { harnessId: 'fake', providerId: 'ghost-provider', modelId: 'ghost-model' };
+// A tuple that is known in the catalog but flagged unavailable.
+const UNAVAILABLE = { harnessId: 'fake', providerId: 'fake-openai', modelId: 'fake-legacy' };
+
+const authGuard: CanActivate = {
+  canActivate(context: ExecutionContext): boolean {
+    const requestContext = context.switchToHttp().getRequest<{ user?: { id: string } }>();
+    requestContext.user = { id: 'user-1' };
+    return true;
+  },
+};
+
+function registryWithFake(): HarnessRegistry {
+  const registry = new HarnessRegistry();
+  registry.register(new FakeHarnessAdapter({ id: 'fake' }));
+  return registry;
+}
+
+describe('Harness selection HTTP surface', () => {
+  let app: INestApplication;
+  let repository: HarnessSelectionRepository;
+
+  beforeAll(async () => {
+    const moduleRef = await Test.createTestingModule({
+      imports: [HarnessModule],
+    })
+      .overrideGuard(AuthGuard)
+      .useValue(authGuard)
+      .overrideProvider(HARNESS_REGISTRY)
+      .useValue(registryWithFake())
+      .compile();
+
+    // Real in-memory repository from the module graph — proves the module wired it.
+    repository = moduleRef.get(HarnessSelectionRepository);
+
+    app = moduleRef.createNestApplication<NestFastifyApplication>(new FastifyAdapter());
+    app.useGlobalPipes(
+      new ValidationPipe({ whitelist: true, forbidNonWhitelisted: true, transform: true }),
+    );
+    await app.init();
+    await app.getHttpAdapter().getInstance().ready();
+  });
+
+  beforeEach(() => {
+    // Reset owner-scoped state between tests via the public API surface.
+    repository.set({ userId: 'user-1', tenantId: 'user-1' }, VALID);
+  });
+
+  afterAll(async () => {
+    await app.close();
+  });
+
+  it('GET selection is server-scoped and ignores caller-supplied scope in the query', async () => {
+    const response = await request(app.getHttpServer())
+      .get('/api/chat/preferences/selection')
+      .query({ userId: 'attacker', tenantId: 'attacker-tenant', seatId: 'attacker-seat' });
+
+    expect(response.status).toBe(200);
+    // The returned selection is user-1's (guard-derived scope), not the query's.
+    expect(response.body.selection).toEqual(VALID);
+  });
+
+  it('PUT with a valid structured tuple persists and round-trips via GET', async () => {
+    const next = { harnessId: 'fake', providerId: 'fake-openai', modelId: 'fake-pro' };
+
+    const put = await request(app.getHttpServer())
+      .put('/api/chat/preferences/selection')
+      .send(next)
+      .set('Content-Type', 'application/json');
+    expect(put.status).toBe(200);
+    expect(put.body.selection).toEqual(next);
+
+    const get = await request(app.getHttpServer()).get('/api/chat/preferences/selection');
+    expect(get.status).toBe(200);
+    expect(get.body.selection).toEqual(next);
+  });
+
+  it('PUT with FREE TEXT is rejected 400 and does not mutate the stored selection', async () => {
+    const response = await request(app.getHttpServer())
+      .put('/api/chat/preferences/selection')
+      .send({ selection: 'gpt-4o' })
+      .set('Content-Type', 'application/json');
+
+    expect(response.status).toBe(400);
+
+    const get = await request(app.getHttpServer()).get('/api/chat/preferences/selection');
+    expect(get.body.selection).toEqual(VALID);
+  });
+
+  it.each([
+    ['seatId', { ...VALID, seatId: 'attacker-seat' }],
+    ['tenantId', { ...VALID, tenantId: 'attacker-tenant' }],
+    ['userId', { ...VALID, userId: 'attacker' }],
+    ['nativeSessionPath', { ...VALID, nativeSessionPath: '/var/native/x.jsonl' }],
+    ['executable', { ...VALID, executable: '/usr/bin/evil' }],
+    ['home', { ...VALID, home: '/home/attacker' }],
+    ['cwd', { ...VALID, cwd: '/tmp/attacker' }],
+  ])(
+    'PUT with an extra authority-bearing field (%s) is rejected 400 and does not mutate stored selection',
+    async (_name, body) => {
+      const response = await request(app.getHttpServer())
+        .put('/api/chat/preferences/selection')
+        .send(body)
+        .set('Content-Type', 'application/json');
+
+      expect(response.status).toBe(400);
+
+      const get = await request(app.getHttpServer()).get('/api/chat/preferences/selection');
+      expect(get.body.selection).toEqual(VALID);
+    },
+  );
+
+  it('PUT with an UNKNOWN tuple returns selection_invalid, unchanged and echoed unchanged (no fallback)', async () => {
+    const response = await request(app.getHttpServer())
+      .put('/api/chat/preferences/selection')
+      .send(UNKNOWN)
+      .set('Content-Type', 'application/json');
+
+    expect(response.status).toBe(422);
+    expect(response.body.code).toBe('selection_invalid');
+    // Echoed back unchanged: no first-row / first-provider substitution.
+    expect(response.body.selection).toEqual(UNKNOWN);
+
+    const get = await request(app.getHttpServer()).get('/api/chat/preferences/selection');
+    expect(get.body.selection).toEqual(VALID);
+  });
+
+  it('PUT with a KNOWN-but-UNAVAILABLE tuple returns model_unavailable, unchanged (distinct from selection_invalid)', async () => {
+    const response = await request(app.getHttpServer())
+      .put('/api/chat/preferences/selection')
+      .send(UNAVAILABLE)
+      .set('Content-Type', 'application/json');
+
+    expect(response.status).toBe(422);
+    expect(response.body.code).toBe('model_unavailable');
+    expect(response.body.selection).toEqual(UNAVAILABLE);
+
+    const get = await request(app.getHttpServer()).get('/api/chat/preferences/selection');
+    expect(get.body.selection).toEqual(VALID);
+  });
+});

apps/gateway/src/harness/harness-selection.controller.ts

@@ -0,0 +1,46 @@
+import { Body, Controller, Get, HttpException, HttpStatus, Put, UseGuards } from '@nestjs/common';
+import { AuthGuard } from '../auth/auth.guard.js';
+import { CurrentUser } from '../auth/current-user.decorator.js';
+import { scopeFromUser, type AuthenticatedUserLike } from '../auth/session-scope.js';
+import { HarnessOperationError } from './harness.registry.js';
+import { HarnessSelectionService } from './harness-selection.service.js';
+import { HarnessSelectionInputDto, type SelectionResponseDto } from './harness.dto.js';
+
+/**
+ * Chat-preferences selection surface. The scope is ALWAYS derived on the server
+ * from the authenticated user (`scopeFromUser(CurrentUser)`); the request body and
+ * query string can never name another user, tenant, or seat. A typed selection
+ * failure (unknown tuple → `selection_invalid`, known-but-unavailable →
+ * `model_unavailable`) is returned as 422 with the requested tuple echoed back
+ * unchanged, and never mutates the stored selection.
+ */
+@Controller('api/chat/preferences/selection')
+@UseGuards(AuthGuard)
+export class HarnessSelectionController {
+  constructor(private readonly selection: HarnessSelectionService) {}
+
+  @Get()
+  get(@CurrentUser() user: AuthenticatedUserLike): SelectionResponseDto {
+    return { selection: this.selection.getSelection(scopeFromUser(user)) };
+  }
+
+  @Put()
+  async put(
+    @CurrentUser() user: AuthenticatedUserLike,
+    @Body() dto: HarnessSelectionInputDto,
+  ): Promise<SelectionResponseDto> {
+    try {
+      const stored = await this.selection.setSelection(scopeFromUser(user), {
+        harnessId: dto.harnessId,
+        providerId: dto.providerId,
+        modelId: dto.modelId,
+      });
+      return { selection: stored };
+    } catch (error) {
+      if (error instanceof HarnessOperationError) {
+        throw new HttpException(error.dto, HttpStatus.UNPROCESSABLE_ENTITY);
+      }
+      throw error;
+    }
+  }
+}

apps/gateway/src/harness/harness-selection.service.ts

@@ -0,0 +1,90 @@
+import { randomUUID } from 'node:crypto';
+import { Inject, Injectable } from '@nestjs/common';
+import type { HarnessSelection } from '@mosaicstack/types';
+import type { ActorTenantScope } from '../auth/session-scope.js';
+import {
+  HarnessAdapterUnavailableError,
+  HarnessRegistry,
+  operationError,
+} from './harness.registry.js';
+import { HARNESS_REGISTRY } from './harness.tokens.js';
+import { readContextFromScope } from './harness.dto.js';
+import { HarnessSelectionRepository } from './harness-selection.repository.js';
+
+/**
+ * Selection logic for the Slice-Zero chat-preferences surface. It validates the
+ * requested harness/provider/model tuple against the live catalog with NO
+ * fallback substitution, then persists it owner-scoped. The stored selection is
+ * only ever mutated when the tuple is valid AND available.
+ */
+@Injectable()
+export class HarnessSelectionService {
+  constructor(
+    @Inject(HARNESS_REGISTRY) private readonly registry: HarnessRegistry,
+    private readonly repository: HarnessSelectionRepository,
+  ) {}
+
+  getSelection(scope: ActorTenantScope): HarnessSelection | null {
+    return this.repository.get(scope);
+  }
+
+  async setSelection(
+    scope: ActorTenantScope,
+    selection: HarnessSelection,
+  ): Promise<HarnessSelection> {
+    // Throws HarnessOperationError (selection_invalid / model_unavailable) with the
+    // requested tuple echoed back unchanged. The store is untouched on any throw.
+    await this.assertSelectionAvailable(scope, selection);
+    return this.repository.set(scope, selection);
+  }
+
+  private async assertSelectionAvailable(
+    scope: ActorTenantScope,
+    selection: HarnessSelection,
+  ): Promise<void> {
+    const correlationId = randomUUID();
+
+    let adapter;
+    try {
+      adapter = this.registry.get(selection.harnessId);
+    } catch (error) {
+      if (error instanceof HarnessAdapterUnavailableError) {
+        // An unknown harness makes the whole tuple invalid — no fallback adapter.
+        throw operationError(
+          'selection_invalid',
+          'The requested harness/provider/model tuple is not in the catalog.',
+          selection,
+          correlationId,
+        );
+      }
+      throw error;
+    }
+
+    const catalog = await adapter.catalog(readContextFromScope(scope));
+    const entry = catalog.models.find(
+      (candidate) =>
+        candidate.harnessId === selection.harnessId &&
+        candidate.providerId === selection.providerId &&
+        candidate.modelId === selection.modelId,
+    );
+
+    if (!entry) {
+      // No first-row / first-provider fallback: reject the requested tuple unchanged.
+      throw operationError(
+        'selection_invalid',
+        'The requested harness/provider/model tuple is not in the catalog.',
+        selection,
+        correlationId,
+      );
+    }
+    if (entry.availability === 'unavailable') {
+      throw operationError(
+        'model_unavailable',
+        'The requested model is currently unavailable.',
+        selection,
+        correlationId,
+        true,
+      );
+    }
+  }
+}

apps/gateway/src/harness/harness.controller.spec.ts

@@ -0,0 +1,138 @@
+import 'reflect-metadata';
+import {
+  type CanActivate,
+  type ExecutionContext,
+  type INestApplication,
+  ValidationPipe,
+} from '@nestjs/common';
+import { FastifyAdapter, type NestFastifyApplication } from '@nestjs/platform-fastify';
+import { Test } from '@nestjs/testing';
+import request from 'supertest';
+import { afterAll, beforeAll, describe, expect, it } from 'vitest';
+import { AuthGuard } from '../auth/auth.guard.js';
+import { HarnessRegistry } from './harness.registry.js';
+import { HARNESS_REGISTRY } from './harness.tokens.js';
+import { FakeHarnessAdapter } from './testing/fake-harness.adapter.js';
+// The real module under test — importing it (not a hand-listed controllers/mocks
+// list) is what makes an unresolved provider fail loudly at app.init() (#1145 guard).
+import { HarnessModule } from './harness.module.js';
+
+// Fields that must NEVER surface on a browser-facing catalog/list response.
+const FORBIDDEN_KEYS = [
+  'executable',
+  'executablePath',
+  'home',
+  'homeDir',
+  'cwd',
+  'workingDir',
+  'workingDirectory',
+  'nativeSessionPath',
+  'sessionPath',
+  'env',
+  'secret',
+  'secrets',
+  'token',
+  'apiKey',
+];
+
+function assertNoForbiddenLeak(payload: unknown): void {
+  const serialized = JSON.stringify(payload).toLowerCase();
+  for (const key of FORBIDDEN_KEYS) {
+    expect(serialized).not.toContain(key.toLowerCase());
+  }
+}
+
+const authGuard: CanActivate = {
+  canActivate(context: ExecutionContext): boolean {
+    const requestContext = context.switchToHttp().getRequest<{ user?: { id: string } }>();
+    requestContext.user = { id: 'user-1' };
+    return true;
+  },
+};
+
+function registryWithFake(): HarnessRegistry {
+  const registry = new HarnessRegistry();
+  registry.register(new FakeHarnessAdapter({ id: 'fake' }));
+  return registry;
+}
+
+describe('Harness catalog HTTP surface', () => {
+  let app: INestApplication;
+
+  beforeAll(async () => {
+    const moduleRef = await Test.createTestingModule({
+      imports: [HarnessModule],
+    })
+      .overrideGuard(AuthGuard)
+      .useValue(authGuard)
+      .overrideProvider(HARNESS_REGISTRY)
+      .useValue(registryWithFake())
+      .compile();
+
+    app = moduleRef.createNestApplication<NestFastifyApplication>(new FastifyAdapter());
+    app.useGlobalPipes(
+      new ValidationPipe({ whitelist: true, forbidNonWhitelisted: true, transform: true }),
+    );
+    await app.init();
+    await app.getHttpAdapter().getInstance().ready();
+  });
+
+  afterAll(async () => {
+    await app.close();
+  });
+
+  it('boots the real HarnessModule so all providers resolve at app.init()', () => {
+    // If HarnessModule failed to resolve a provider, beforeAll's app.init() would
+    // have thrown and this suite would never reach here.
+    expect(app).toBeDefined();
+  });
+
+  it('GET /api/harnesses returns 200 with safe fields only', async () => {
+    const response = await request(app.getHttpServer()).get('/api/harnesses');
+
+    expect(response.status).toBe(200);
+    expect(Array.isArray(response.body)).toBe(true);
+    expect(response.body.length).toBeGreaterThan(0);
+    const summary = response.body[0];
+    expect(Object.keys(summary).sort()).toEqual(['capabilities', 'displayName', 'id']);
+    expect(summary.id).toBe('fake');
+    expect(typeof summary.displayName).toBe('string');
+    expect(Array.isArray(summary.capabilities)).toBe(true);
+    assertNoForbiddenLeak(response.body);
+  });
+
+  it('GET /api/harnesses/:harnessId/catalog returns 200 with safe catalog fields only', async () => {
+    const response = await request(app.getHttpServer()).get('/api/harnesses/fake/catalog');
+
+    expect(response.status).toBe(200);
+    expect(response.body.harnessId).toBe('fake');
+    expect(typeof response.body.version).toBe('string');
+    expect(typeof response.body.fingerprint).toBe('string');
+    expect(Array.isArray(response.body.models)).toBe(true);
+    expect(response.body.models.length).toBeGreaterThan(0);
+    const entry = response.body.models[0];
+    // Whitelisted catalog-entry fields only (no executables/paths/secrets).
+    expect(Object.keys(entry).sort()).toEqual(
+      [
+        'authState',
+        'availability',
+        'displayName',
+        'harnessId',
+        'inputTypes',
+        'modelId',
+        'providerId',
+        'reasoningCapability',
+      ].sort(),
+    );
+    assertNoForbiddenLeak(response.body);
+  });
+
+  it('GET catalog for an unknown harnessId returns a typed adapter_unavailable error, never a fallback catalog', async () => {
+    const response = await request(app.getHttpServer()).get('/api/harnesses/ghost-harness/catalog');
+
+    expect(response.status).toBe(404);
+    expect(response.body.code).toBe('adapter_unavailable');
+    // A fallback catalog would carry a models array; a typed error must not.
+    expect(response.body.models).toBeUndefined();
+  });
+});

apps/gateway/src/harness/harness.controller.ts

@@ -0,0 +1,65 @@
+import {
+  Controller,
+  Get,
+  HttpException,
+  HttpStatus,
+  Inject,
+  Param,
+  UseGuards,
+} from '@nestjs/common';
+import { AuthGuard } from '../auth/auth.guard.js';
+import { CurrentUser } from '../auth/current-user.decorator.js';
+import { scopeFromUser, type AuthenticatedUserLike } from '../auth/session-scope.js';
+import { HarnessAdapterUnavailableError, HarnessRegistry } from './harness.registry.js';
+import { HARNESS_REGISTRY } from './harness.tokens.js';
+import {
+  readContextFromScope,
+  toHarnessSummary,
+  toSafeCatalog,
+  type HarnessCatalogDto,
+  type HarnessSummaryDto,
+} from './harness.dto.js';
+
+/**
+ * Generic harness catalog surface. It exposes only harness-neutral, browser-safe
+ * fields (identity, capabilities, provider/model catalog) — never executables,
+ * native paths, home/cwd, env, or secrets. There is NO provider-probe route here;
+ * `/api/providers` and `POST /api/providers/test` are intentionally out of scope.
+ */
+@Controller('api/harnesses')
+@UseGuards(AuthGuard)
+export class HarnessController {
+  constructor(@Inject(HARNESS_REGISTRY) private readonly registry: HarnessRegistry) {}
+
+  @Get()
+  async list(@CurrentUser() user: AuthenticatedUserLike): Promise<HarnessSummaryDto[]> {
+    const context = readContextFromScope(scopeFromUser(user));
+    const summaries: HarnessSummaryDto[] = [];
+    for (const adapter of this.registry.list()) {
+      summaries.push(toHarnessSummary(await adapter.describe(context)));
+    }
+    return summaries;
+  }
+
+  @Get(':harnessId/catalog')
+  async catalog(
+    @CurrentUser() user: AuthenticatedUserLike,
+    @Param('harnessId') harnessId: string,
+  ): Promise<HarnessCatalogDto> {
+    const context = readContextFromScope(scopeFromUser(user));
+    let adapter;
+    try {
+      adapter = this.registry.get(harnessId);
+    } catch (error) {
+      if (error instanceof HarnessAdapterUnavailableError) {
+        // Typed failure — NEVER a fallback catalog for an unknown harness id.
+        throw new HttpException(
+          { code: error.code, message: error.message, harnessId },
+          HttpStatus.NOT_FOUND,
+        );
+      }
+      throw error;
+    }
+    return toSafeCatalog(await adapter.catalog(context));
+  }
+}

apps/gateway/src/harness/harness.dto.ts

@@ -0,0 +1,116 @@
+import { randomUUID } from 'node:crypto';
+import { IsNotEmpty, IsString } from 'class-validator';
+import type {
+  HarnessActorContext,
+  HarnessAuthState,
+  HarnessCapability,
+  HarnessCatalog,
+  HarnessCatalogEntry,
+  HarnessDescriptor,
+  HarnessInputType,
+  HarnessModelAvailability,
+  HarnessSelection,
+} from '@mosaicstack/types';
+import type { ActorTenantScope } from '../auth/session-scope.js';
+
+/**
+ * Structured selection tuple accepted on `PUT /api/chat/preferences/selection`.
+ *
+ * The body is a STRUCTURED tuple (harness + provider + model), never a free-text
+ * model string. With `ValidationPipe({ whitelist: true, forbidNonWhitelisted: true })`
+ * any extra property — including smuggled server-authority fields such as
+ * `seatId`, `tenantId`, `userId`, `nativeSessionPath`, `executable`, `home`, `cwd` —
+ * is rejected with 400. There is deliberately no field through which a caller can
+ * name a scope; scope is derived on the server from the authenticated session.
+ */
+export class HarnessSelectionInputDto {
+  @IsString()
+  @IsNotEmpty()
+  harnessId!: string;
+
+  @IsString()
+  @IsNotEmpty()
+  providerId!: string;
+
+  @IsString()
+  @IsNotEmpty()
+  modelId!: string;
+}
+
+/** Browser-safe harness summary — identity and capabilities only. */
+export interface HarnessSummaryDto {
+  readonly id: string;
+  readonly displayName: string;
+  readonly capabilities: readonly HarnessCapability[];
+}
+
+/** Browser-safe catalog entry — no executables, paths, secrets, or env. */
+export interface HarnessCatalogEntryDto {
+  readonly harnessId: string;
+  readonly providerId: string;
+  readonly modelId: string;
+  readonly displayName: string;
+  readonly reasoningCapability: boolean;
+  readonly inputTypes: readonly HarnessInputType[];
+  readonly authState: HarnessAuthState;
+  readonly availability: HarnessModelAvailability;
+}
+
+/** Browser-safe catalog envelope. */
+export interface HarnessCatalogDto {
+  readonly harnessId: string;
+  readonly version: string;
+  readonly fingerprint: string;
+  readonly models: readonly HarnessCatalogEntryDto[];
+}
+
+/** Response envelope for the caller's current selection (null when unset). */
+export interface SelectionResponseDto {
+  readonly selection: HarnessSelection | null;
+}
+
+/**
+ * Derive a server-trusted {@link HarnessActorContext} for read operations from the
+ * session-derived {@link ActorTenantScope}. All authority originates on the server;
+ * nothing here is caller-supplied. A fresh correlation id is minted per call.
+ */
+export function readContextFromScope(scope: ActorTenantScope): HarnessActorContext {
+  return {
+    actorId: scope.userId,
+    tenantId: scope.tenantId,
+    seatId: scope.userId,
+    correlationId: randomUUID(),
+  };
+}
+
+/** Project a descriptor onto the browser-safe summary shape (whitelist by construction). */
+export function toHarnessSummary(descriptor: HarnessDescriptor): HarnessSummaryDto {
+  return {
+    id: descriptor.id,
+    displayName: descriptor.displayName,
+    capabilities: [...descriptor.capabilities],
+  };
+}
+
+/** Project a catalog onto the browser-safe shape (whitelist by construction). */
+export function toSafeCatalog(catalog: HarnessCatalog): HarnessCatalogDto {
+  return {
+    harnessId: catalog.harnessId,
+    version: catalog.version,
+    fingerprint: catalog.fingerprint,
+    models: catalog.models.map(toSafeCatalogEntry),
+  };
+}
+
+function toSafeCatalogEntry(entry: HarnessCatalogEntry): HarnessCatalogEntryDto {
+  return {
+    harnessId: entry.harnessId,
+    providerId: entry.providerId,
+    modelId: entry.modelId,
+    displayName: entry.displayName,
+    reasoningCapability: entry.reasoningCapability,
+    inputTypes: [...entry.inputTypes],
+    authState: entry.authState,
+    availability: entry.availability,
+  };
+}

apps/gateway/src/harness/harness.module.ts

@@ -0,0 +1,37 @@
+import { Module } from '@nestjs/common';
+import { HarnessRegistry } from './harness.registry.js';
+import { HarnessService } from './harness.service.js';
+import {
+  HARNESS_CONVERSATION_SERVICE,
+  HARNESS_CONVERSATION_SERVICE_UNAVAILABLE,
+  HARNESS_REGISTRY,
+  HARNESS_SERVICE,
+} from './harness.tokens.js';
+import { HarnessController } from './harness.controller.js';
+import { HarnessSelectionController } from './harness-selection.controller.js';
+import { HarnessSelectionService } from './harness-selection.service.js';
+import { HarnessSelectionRepository } from './harness-selection.repository.js';
+
+/**
+ * Wires the harness-neutral registry/service (Task Two) together with the
+ * Slice-Zero catalog and selection HTTP surfaces (Task Three).
+ *
+ * The registry is provided empty here; real harness adapters are registered in a
+ * later task. Because the controllers/services resolve their collaborators through
+ * this real module graph, an unresolved provider fails loudly at `app.init()`.
+ */
+@Module({
+  controllers: [HarnessController, HarnessSelectionController],
+  providers: [
+    { provide: HARNESS_REGISTRY, useFactory: () => new HarnessRegistry() },
+    { provide: HARNESS_SERVICE, useClass: HarnessService },
+    // Task Five: bind the conversation-service token to its explicit "not yet bound"
+    // sentinel. The pi-rpc router treats this as a hard, typed startup failure; Task 14
+    // replaces it with a real service. Exported so ChatModule's router can inject it.
+    { provide: HARNESS_CONVERSATION_SERVICE, useValue: HARNESS_CONVERSATION_SERVICE_UNAVAILABLE },
+    HarnessSelectionRepository,
+    HarnessSelectionService,
+  ],
+  exports: [HARNESS_REGISTRY, HARNESS_SERVICE, HARNESS_CONVERSATION_SERVICE],
+})
+export class HarnessModule {}

apps/gateway/src/harness/harness.registry.spec.ts

@@ -0,0 +1,69 @@
+import { describe, expect, it } from 'vitest';
+import {
+  HarnessAdapterUnavailableError,
+  HarnessRegistrationError,
+  HarnessRegistry,
+} from './harness.registry.js';
+import { FakeHarnessAdapter } from './testing/fake-harness.adapter.js';
+
+describe('HarnessRegistry', () => {
+  it('registers and looks up an adapter by harness id', () => {
+    const registry = new HarnessRegistry();
+    const adapter = new FakeHarnessAdapter({ id: 'fake' });
+
+    registry.register(adapter);
+
+    expect(registry.get('fake')).toBe(adapter);
+    expect(registry.has('fake')).toBe(true);
+    expect(registry.list().map((entry) => entry.id)).toEqual(['fake']);
+  });
+
+  it('rejects a blank adapter id', () => {
+    const registry = new HarnessRegistry();
+
+    let error: unknown;
+    try {
+      registry.register(new FakeHarnessAdapter({ id: '   ' }));
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessRegistrationError);
+    expect((error as HarnessRegistrationError).reason).toBe('blank_id');
+    expect(registry.list()).toEqual([]);
+  });
+
+  it('rejects a duplicate adapter id', () => {
+    const registry = new HarnessRegistry();
+    registry.register(new FakeHarnessAdapter({ id: 'fake' }));
+
+    let error: unknown;
+    try {
+      registry.register(new FakeHarnessAdapter({ id: 'fake' }));
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessRegistrationError);
+    expect((error as HarnessRegistrationError).reason).toBe('duplicate_id');
+    expect((error as HarnessRegistrationError).harnessId).toBe('fake');
+    // The original registration is untouched.
+    expect(registry.list()).toHaveLength(1);
+  });
+
+  it('returns adapter_unavailable for an unknown harness id', () => {
+    const registry = new HarnessRegistry();
+
+    let error: unknown;
+    try {
+      registry.get('missing');
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessAdapterUnavailableError);
+    expect((error as HarnessAdapterUnavailableError).code).toBe('adapter_unavailable');
+    expect((error as HarnessAdapterUnavailableError).harnessId).toBe('missing');
+    expect(registry.has('missing')).toBe(false);
+  });
+});

apps/gateway/src/harness/harness.registry.ts

@@ -0,0 +1,100 @@
+import { Injectable } from '@nestjs/common';
+import type {
+  HarnessAdapter,
+  HarnessErrorCode,
+  HarnessErrorDto,
+  HarnessSelection,
+} from '@mosaicstack/types';
+
+/**
+ * A typed harness operation failure that carries a fully-formed, browser-safe
+ * {@link HarnessErrorDto}. The DTO's `selection` is always the exact requested
+ * tuple — there is no field through which a substituted "effective" selection
+ * could ever be reported.
+ */
+export class HarnessOperationError extends Error {
+  readonly code: HarnessErrorCode;
+  readonly dto: HarnessErrorDto;
+
+  constructor(dto: HarnessErrorDto) {
+    super(dto.message);
+    this.name = 'HarnessOperationError';
+    this.code = dto.code;
+    this.dto = dto;
+  }
+}
+
+/** Build a {@link HarnessOperationError} that echoes the requested selection unchanged. */
+export function operationError(
+  code: HarnessErrorCode,
+  message: string,
+  selection: HarnessSelection,
+  correlationId: string,
+  retryable = false,
+): HarnessOperationError {
+  return new HarnessOperationError({ code, message, retryable, correlationId, selection });
+}
+
+/** Raised when an unknown harness id is looked up. Discriminated by `code`. */
+export class HarnessAdapterUnavailableError extends Error {
+  readonly code = 'adapter_unavailable' as const satisfies HarnessErrorCode;
+
+  constructor(readonly harnessId: string) {
+    super(`No harness adapter is registered for id "${harnessId}".`);
+    this.name = 'HarnessAdapterUnavailableError';
+  }
+}
+
+export type HarnessRegistrationFailure = 'blank_id' | 'duplicate_id';
+
+/** Raised when an adapter cannot be registered (blank or duplicate id). */
+export class HarnessRegistrationError extends Error {
+  constructor(
+    readonly reason: HarnessRegistrationFailure,
+    readonly harnessId: string,
+  ) {
+    super(
+      reason === 'blank_id'
+        ? 'A harness adapter id must be a non-empty string.'
+        : `A harness adapter is already registered for id "${harnessId}".`,
+    );
+    this.name = 'HarnessRegistrationError';
+  }
+}
+
+/**
+ * Harness-neutral adapter registry. Adapters are keyed by their harness id.
+ * Registration rejects blank and duplicate ids; lookup of an unknown id fails
+ * with {@link HarnessAdapterUnavailableError} (`adapter_unavailable`).
+ */
+@Injectable()
+export class HarnessRegistry {
+  private readonly adapters = new Map<string, HarnessAdapter>();
+
+  register(adapter: HarnessAdapter): void {
+    const id = adapter.id;
+    if (typeof id !== 'string' || id.trim().length === 0) {
+      throw new HarnessRegistrationError('blank_id', id ?? '');
+    }
+    if (this.adapters.has(id)) {
+      throw new HarnessRegistrationError('duplicate_id', id);
+    }
+    this.adapters.set(id, adapter);
+  }
+
+  get(harnessId: string): HarnessAdapter {
+    const adapter = this.adapters.get(harnessId);
+    if (!adapter) {
+      throw new HarnessAdapterUnavailableError(harnessId);
+    }
+    return adapter;
+  }
+
+  has(harnessId: string): boolean {
+    return this.adapters.has(harnessId);
+  }
+
+  list(): readonly HarnessAdapter[] {
+    return [...this.adapters.values()];
+  }
+}

apps/gateway/src/harness/harness.service.spec.ts

@@ -0,0 +1,227 @@
+import { describe, expect, it } from 'vitest';
+import type { HarnessActorContext, HarnessCapability, HarnessSelection } from '@mosaicstack/types';
+import { HARNESS_CAPABILITIES } from '@mosaicstack/types';
+import { HarnessOperationError, HarnessRegistry } from './harness.registry.js';
+import {
+  HarnessScopeViolationError,
+  HarnessService,
+  type TrustedGatewayScope,
+} from './harness.service.js';
+import { FakeHarnessAdapter } from './testing/fake-harness.adapter.js';
+
+const SCOPE: TrustedGatewayScope = {
+  actorId: 'actor-trusted',
+  tenantId: 'tenant-trusted',
+  seatId: 'seat-trusted',
+  correlationId: 'correlation-trusted',
+};
+
+const READ_CONTEXT: HarnessActorContext = {
+  actorId: SCOPE.actorId,
+  tenantId: SCOPE.tenantId,
+  seatId: SCOPE.seatId,
+  correlationId: SCOPE.correlationId,
+};
+
+function setup(capabilities?: readonly HarnessCapability[]) {
+  const registry = new HarnessRegistry();
+  const adapter = new FakeHarnessAdapter({ id: 'fake', capabilities });
+  registry.register(adapter);
+  const service = new HarnessService(registry);
+  return { registry, adapter, service };
+}
+
+async function availableSelection(adapter: FakeHarnessAdapter): Promise<HarnessSelection> {
+  const catalog = await adapter.catalog(READ_CONTEXT);
+  const entry = catalog.models.find((model) => model.availability === 'available');
+  if (!entry) {
+    throw new Error('fixture requires an available model');
+  }
+  return { harnessId: entry.harnessId, providerId: entry.providerId, modelId: entry.modelId };
+}
+
+describe('HarnessService', () => {
+  it('derives the actor context from trusted scope on create', async () => {
+    const { service, adapter } = setup();
+    const selection = await availableSelection(adapter);
+
+    const snapshot = await service.createSession(SCOPE, {
+      conversationId: 'conversation-1',
+      selection,
+    });
+
+    expect(snapshot.seatId).toBe(SCOPE.seatId);
+    expect(snapshot.state).toBe('idle');
+    expect(snapshot.selection).toEqual(selection);
+    expect(snapshot.nativeSessionId).toBeTruthy();
+  });
+
+  it('rejects server-authority fields supplied by an external caller', async () => {
+    const { service, adapter } = setup();
+    const selection = await availableSelection(adapter);
+
+    const hostile = {
+      conversationId: 'conversation-1',
+      selection,
+      seatId: 'attacker-seat',
+      executablePath: '/usr/bin/evil',
+      home: '/home/attacker',
+      cwd: '/tmp/attacker',
+      nativeSessionPath: '/var/native/attacker.jsonl',
+    } as unknown as Parameters<HarnessService['createSession']>[1];
+
+    let error: unknown;
+    try {
+      await service.createSession(SCOPE, hostile);
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessScopeViolationError);
+    expect((error as HarnessScopeViolationError).field).toBe('seatId');
+  });
+
+  it('returns adapter_unavailable for an unknown harness id, echoing the requested tuple', async () => {
+    const { service } = setup();
+    const selection: HarnessSelection = {
+      harnessId: 'ghost-harness',
+      providerId: 'p',
+      modelId: 'm',
+    };
+
+    let error: unknown;
+    try {
+      await service.createSession(SCOPE, { conversationId: 'conversation-1', selection });
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessOperationError);
+    const dto = (error as HarnessOperationError).dto;
+    expect(dto.code).toBe('adapter_unavailable');
+    expect(dto.selection).toEqual(selection);
+    expect(dto.correlationId).toBe(SCOPE.correlationId);
+  });
+
+  it('returns selection_invalid for an unknown provider/model tuple, unchanged', async () => {
+    const { service } = setup();
+    const selection: HarnessSelection = {
+      harnessId: 'fake',
+      providerId: 'ghost-provider',
+      modelId: 'ghost-model',
+    };
+
+    let error: unknown;
+    try {
+      await service.createSession(SCOPE, { conversationId: 'conversation-1', selection });
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessOperationError);
+    const dto = (error as HarnessOperationError).dto;
+    expect(dto.code).toBe('selection_invalid');
+    expect(dto.selection).toEqual(selection);
+  });
+
+  it('returns model_unavailable without falling back for a known unavailable model', async () => {
+    const { service, adapter } = setup();
+    const catalog = await adapter.catalog(READ_CONTEXT);
+    const unavailable = catalog.models.find((entry) => entry.availability === 'unavailable');
+    expect(unavailable).toBeDefined();
+    const selection: HarnessSelection = {
+      harnessId: unavailable!.harnessId,
+      providerId: unavailable!.providerId,
+      modelId: unavailable!.modelId,
+    };
+
+    let error: unknown;
+    try {
+      await service.createSession(SCOPE, { conversationId: 'conversation-1', selection });
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessOperationError);
+    const dto = (error as HarnessOperationError).dto;
+    expect(dto.code).toBe('model_unavailable');
+    // No substitution: the DTO tuple is exactly what was requested.
+    expect(dto.selection).toEqual(selection);
+  });
+
+  it('gives create, resume, detach, evict, and end distinct observable effects', async () => {
+    const { service, adapter } = setup();
+    const selection = await availableSelection(adapter);
+
+    const created = await service.createSession(SCOPE, {
+      conversationId: 'conversation-create',
+      selection,
+    });
+    expect(created.state).toBe('idle');
+    expect(created.processId).toBeTruthy();
+    expect(created.attachedClientIds).toEqual([]);
+
+    const resumed = await service.resumeSession(SCOPE, {
+      conversationId: 'conversation-resume',
+      nativeSessionId: 'native-preexisting-123',
+      selection,
+    });
+    // Resume binds the supplied native session; create mints a fresh one.
+    expect(resumed.nativeSessionId).toBe('native-preexisting-123');
+    expect(resumed.nativeSessionId).not.toBe(created.nativeSessionId);
+
+    await service.attach(SCOPE, {
+      conversationId: 'conversation-create',
+      clientId: 'browser-1',
+    });
+    const afterAttach = await service.snapshot(SCOPE, 'conversation-create');
+    expect(afterAttach.attachedClientIds).toEqual(['browser-1']);
+
+    const afterDetach = await service.detach(SCOPE, {
+      conversationId: 'conversation-create',
+      clientId: 'browser-1',
+    });
+    // Detach removes the browser attachment only; the process stays alive.
+    expect(afterDetach.attachedClientIds).toEqual([]);
+    expect(afterDetach.state).toBe('idle');
+    expect(afterDetach.processId).toBeTruthy();
+
+    const afterEvict = await service.evict(SCOPE, {
+      conversationId: 'conversation-create',
+      reason: 'idle_timeout',
+    });
+    // Evict stops the process but retains the resumable native session.
+    expect(afterEvict.state).toBe('evicted');
+    expect(afterEvict.processId).toBeUndefined();
+    expect(afterEvict.nativeSessionId).toBe(created.nativeSessionId);
+
+    const afterEnd = await service.end(SCOPE, {
+      conversationId: 'conversation-create',
+      reason: 'session_ended',
+    });
+    // End destructively terminates the native session.
+    expect(afterEnd.state).toBe('ended');
+  });
+
+  it('fails typed when an unsupported capability is exercised', async () => {
+    const withoutExtensionUi = HARNESS_CAPABILITIES.filter(
+      (capability) => capability !== 'extensionUi',
+    );
+    const { service, adapter } = setup(withoutExtensionUi);
+    const selection = await availableSelection(adapter);
+    await service.createSession(SCOPE, { conversationId: 'conversation-1', selection });
+
+    let error: unknown;
+    try {
+      await service.respondInteraction(SCOPE, {
+        conversationId: 'conversation-1',
+        response: { requestId: 'interaction-1', type: 'confirm', accepted: true },
+      });
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessOperationError);
+    expect((error as HarnessOperationError).dto.code).toBe('interaction_unsupported');
+  });
+});

apps/gateway/src/harness/harness.service.ts

@@ -0,0 +1,285 @@
+import { Inject, Injectable } from '@nestjs/common';
+import type {
+  HarnessActorContext,
+  HarnessAdapter,
+  HarnessCatalog,
+  HarnessCloseReason,
+  HarnessInteractionResponse,
+  HarnessSelection,
+  HarnessSessionHandle,
+  HarnessSessionSnapshot,
+} from '@mosaicstack/types';
+import {
+  HarnessAdapterUnavailableError,
+  HarnessRegistry,
+  operationError,
+} from './harness.registry.js';
+import { HARNESS_REGISTRY } from './harness.tokens.js';
+
+/**
+ * Trusted, server-derived authority. In production this is produced by the
+ * Gateway from the authenticated session — never from a browser/caller DTO.
+ */
+export interface TrustedGatewayScope {
+  readonly actorId: string;
+  readonly tenantId: string;
+  readonly seatId: string;
+  readonly correlationId: string;
+}
+
+/** Server-authority fields that must never arrive from an external request DTO. */
+const FORBIDDEN_REQUEST_FIELDS = [
+  'actorId',
+  'tenantId',
+  'correlationId',
+  'seatId',
+  'seat',
+  'executable',
+  'executablePath',
+  'home',
+  'homeDir',
+  'cwd',
+  'workingDir',
+  'workingDirectory',
+  'nativeSessionPath',
+  'sessionPath',
+] as const;
+
+/** Raised when an external request DTO smuggles a server-authority field. */
+export class HarnessScopeViolationError extends Error {
+  constructor(readonly field: string) {
+    super(`External request supplied server-authority field "${field}".`);
+    this.name = 'HarnessScopeViolationError';
+  }
+}
+
+export interface CreateHarnessSessionRequest {
+  readonly conversationId: string;
+  readonly selection: HarnessSelection;
+}
+
+export interface ResumeHarnessSessionRequest {
+  readonly conversationId: string;
+  readonly nativeSessionId: string;
+  readonly selection: HarnessSelection;
+}
+
+export interface AttachClientRequest {
+  readonly conversationId: string;
+  readonly clientId: string;
+}
+
+export interface DetachClientRequest {
+  readonly conversationId: string;
+  readonly clientId: string;
+}
+
+export interface EvictSessionRequest {
+  readonly conversationId: string;
+  readonly reason: HarnessCloseReason;
+}
+
+export interface EndSessionRequest {
+  readonly conversationId: string;
+  readonly reason: HarnessCloseReason;
+}
+
+export interface RespondInteractionRequest {
+  readonly conversationId: string;
+  readonly response: HarnessInteractionResponse;
+}
+
+interface ActiveSession {
+  readonly harnessId: string;
+  readonly handle: HarnessSessionHandle;
+  readonly correlationId: string;
+}
+
+/**
+ * Harness-neutral service. It derives the {@link HarnessActorContext} strictly
+ * from trusted Gateway scope, validates the selected provider/model tuple with
+ * NO fallback substitution, and exposes distinct create/resume/detach/evict/end
+ * lifecycle operations.
+ */
+@Injectable()
+export class HarnessService {
+  private readonly sessions = new Map<string, ActiveSession>();
+
+  constructor(@Inject(HARNESS_REGISTRY) private readonly registry: HarnessRegistry) {}
+
+  async createSession(
+    scope: TrustedGatewayScope,
+    request: CreateHarnessSessionRequest,
+  ): Promise<HarnessSessionSnapshot> {
+    assertTrustedRequest(request);
+    const { conversationId, selection } = request;
+    const adapter = this.resolveAdapter(scope, selection);
+    const context = deriveActorContext(scope);
+    await this.assertSelectionAvailable(scope, adapter.catalog(context), selection);
+
+    const handle = await adapter.create({ context, conversationId, selection });
+    this.sessions.set(conversationId, {
+      harnessId: selection.harnessId,
+      handle,
+      correlationId: scope.correlationId,
+    });
+    return handle.snapshot();
+  }
+
+  async resumeSession(
+    scope: TrustedGatewayScope,
+    request: ResumeHarnessSessionRequest,
+  ): Promise<HarnessSessionSnapshot> {
+    assertTrustedRequest(request);
+    const { conversationId, nativeSessionId, selection } = request;
+    const adapter = this.resolveAdapter(scope, selection);
+    const context = deriveActorContext(scope);
+    await this.assertSelectionAvailable(scope, adapter.catalog(context), selection);
+
+    const handle = await adapter.resume({ context, conversationId, nativeSessionId, selection });
+    this.sessions.set(conversationId, {
+      harnessId: selection.harnessId,
+      handle,
+      correlationId: scope.correlationId,
+    });
+    return handle.snapshot();
+  }
+
+  async attach(
+    scope: TrustedGatewayScope,
+    request: AttachClientRequest,
+  ): Promise<HarnessSessionSnapshot> {
+    assertTrustedRequest(request);
+    const handle = this.requireHandle(scope, request.conversationId);
+    await handle.attach({ clientId: request.clientId });
+    return handle.snapshot();
+  }
+
+  async detach(
+    scope: TrustedGatewayScope,
+    request: DetachClientRequest,
+  ): Promise<HarnessSessionSnapshot> {
+    assertTrustedRequest(request);
+    const handle = this.requireHandle(scope, request.conversationId);
+    await handle.detach(request.clientId);
+    return handle.snapshot();
+  }
+
+  async evict(
+    scope: TrustedGatewayScope,
+    request: EvictSessionRequest,
+  ): Promise<HarnessSessionSnapshot> {
+    assertTrustedRequest(request);
+    const handle = this.requireHandle(scope, request.conversationId);
+    await handle.evictProcess(request.reason);
+    return handle.snapshot();
+  }
+
+  async end(
+    scope: TrustedGatewayScope,
+    request: EndSessionRequest,
+  ): Promise<HarnessSessionSnapshot> {
+    assertTrustedRequest(request);
+    const handle = this.requireHandle(scope, request.conversationId);
+    await handle.endSession(request.reason);
+    const snapshot = await handle.snapshot();
+    this.sessions.delete(request.conversationId);
+    return snapshot;
+  }
+
+  async respondInteraction(
+    scope: TrustedGatewayScope,
+    request: RespondInteractionRequest,
+  ): Promise<void> {
+    assertTrustedRequest(request);
+    const handle = this.requireHandle(scope, request.conversationId);
+    await handle.respondInteraction(request.response);
+  }
+
+  async snapshot(
+    scope: TrustedGatewayScope,
+    conversationId: string,
+  ): Promise<HarnessSessionSnapshot> {
+    const handle = this.requireHandle(scope, conversationId);
+    return handle.snapshot();
+  }
+
+  private resolveAdapter(scope: TrustedGatewayScope, selection: HarnessSelection): HarnessAdapter {
+    try {
+      return this.registry.get(selection.harnessId);
+    } catch (error) {
+      if (error instanceof HarnessAdapterUnavailableError) {
+        throw operationError('adapter_unavailable', error.message, selection, scope.correlationId);
+      }
+      throw error;
+    }
+  }
+
+  private async assertSelectionAvailable(
+    scope: TrustedGatewayScope,
+    catalogPromise: Promise<HarnessCatalog>,
+    selection: HarnessSelection,
+  ): Promise<void> {
+    const catalog = await catalogPromise;
+    const entry = catalog.models.find(
+      (candidate) =>
+        candidate.harnessId === selection.harnessId &&
+        candidate.providerId === selection.providerId &&
+        candidate.modelId === selection.modelId,
+    );
+    if (!entry) {
+      // No first-row fallback: reject the requested tuple unchanged.
+      throw operationError(
+        'selection_invalid',
+        'The requested harness/provider/model tuple is not in the catalog.',
+        selection,
+        scope.correlationId,
+      );
+    }
+    if (entry.availability === 'unavailable') {
+      throw operationError(
+        'model_unavailable',
+        'The requested model is currently unavailable.',
+        selection,
+        scope.correlationId,
+        true,
+      );
+    }
+  }
+
+  private requireHandle(scope: TrustedGatewayScope, conversationId: string): HarnessSessionHandle {
+    const active = this.sessions.get(conversationId);
+    if (!active) {
+      throw operationError(
+        'session_not_found',
+        `No active harness session for conversation "${conversationId}".`,
+        { harnessId: '', providerId: '', modelId: '' },
+        scope.correlationId,
+      );
+    }
+    return active.handle;
+  }
+}
+
+/** Build the actor context strictly from trusted scope. No caller data leaks in. */
+export function deriveActorContext(scope: TrustedGatewayScope): HarnessActorContext {
+  return {
+    actorId: scope.actorId,
+    tenantId: scope.tenantId,
+    seatId: scope.seatId,
+    correlationId: scope.correlationId,
+  };
+}
+
+/** Reject any request object that carries a server-authority field. */
+function assertTrustedRequest(request: object): void {
+  for (const field of FORBIDDEN_REQUEST_FIELDS) {
+    if (Object.prototype.hasOwnProperty.call(request, field)) {
+      throw new HarnessScopeViolationError(field);
+    }
+  }
+}
+
+// Re-export the typed operation error so callers importing from the service
+// have the discriminated failure type without reaching into the registry.
+export { HarnessOperationError } from './harness.registry.js';

apps/gateway/src/harness/harness.tokens.ts

@@ -0,0 +1,45 @@
+/**
+ * Nest dependency-injection tokens for the harness-neutral registry and service.
+ *
+ * String tokens follow the existing Gateway convention (see `memory/memory.tokens.ts`)
+ * and remain valid Nest `InjectionToken`s for `@Inject(...)`.
+ */
+import type { HarnessConversationService } from '@mosaicstack/types';
+
+export const HARNESS_REGISTRY = 'HARNESS_REGISTRY' as const;
+export const HARNESS_SERVICE = 'HARNESS_SERVICE' as const;
+
+export type HarnessRegistryToken = typeof HARNESS_REGISTRY;
+export type HarnessServiceToken = typeof HARNESS_SERVICE;
+
+/**
+ * Token for the {@link HarnessConversationService} that {@link HarnessChatRuntime}
+ * depends on. Until Task 14 provides a real implementation, `HarnessModule` binds
+ * the {@link HARNESS_CONVERSATION_SERVICE_UNAVAILABLE} sentinel here, and the
+ * `pi-rpc` router treats that sentinel as a hard, typed startup failure.
+ */
+export const HARNESS_CONVERSATION_SERVICE = 'HARNESS_CONVERSATION_SERVICE' as const;
+
+export type HarnessConversationServiceToken = typeof HARNESS_CONVERSATION_SERVICE;
+
+/**
+ * Explicit "not yet bound" value for {@link HARNESS_CONVERSATION_SERVICE}. It is a
+ * distinct sentinel — never `null`/`undefined` — so an unbound service is an
+ * intentional, checkable state rather than an accidental nil that could read as
+ * "present". Replaced by a real service in Task 14.
+ */
+export const HARNESS_CONVERSATION_SERVICE_UNAVAILABLE: unique symbol = Symbol(
+  'HARNESS_CONVERSATION_SERVICE_UNAVAILABLE',
+);
+
+/** A binding for {@link HARNESS_CONVERSATION_SERVICE}: a real service or the sentinel. */
+export type HarnessConversationServiceBinding =
+  | HarnessConversationService
+  | typeof HARNESS_CONVERSATION_SERVICE_UNAVAILABLE;
+
+/** Narrows a binding to a usable service, excluding the unavailable sentinel. */
+export function isHarnessConversationServiceAvailable(
+  binding: HarnessConversationServiceBinding,
+): binding is HarnessConversationService {
+  return binding !== HARNESS_CONVERSATION_SERVICE_UNAVAILABLE;
+}

apps/gateway/src/harness/testing/fake-harness.adapter.spec.ts

@@ -0,0 +1,107 @@
+import { describe, expect, it } from 'vitest';
+import type { HarnessActorContext, HarnessSelection } from '@mosaicstack/types';
+import { HarnessOperationError } from '../harness.registry.js';
+import { FakeHarnessAdapter } from './fake-harness.adapter.js';
+import { runHarnessAdapterContract } from './harness-adapter.contract.js';
+
+const CONTEXT: HarnessActorContext = {
+  actorId: 'actor-1',
+  tenantId: 'tenant-1',
+  seatId: 'seat-1',
+  correlationId: 'correlation-1',
+};
+
+// The reusable conformance suite. Task 13 re-runs it against the native Pi adapter.
+runHarnessAdapterContract('FakeHarnessAdapter', () => new FakeHarnessAdapter({ id: 'fake' }));
+
+describe('FakeHarnessAdapter no-substitution', () => {
+  it('never substitutes the first catalog row when a bogus selection is requested', async () => {
+    const adapter = new FakeHarnessAdapter({ id: 'fake' });
+    const catalog = await adapter.catalog(CONTEXT);
+    const firstRow = catalog.models[0];
+    if (!firstRow) {
+      throw new Error('fixture requires a catalog model');
+    }
+    const available = catalog.models.find(
+      (entry) => entry.availability === 'available' && entry.modelId !== firstRow.modelId,
+    );
+    expect(available).toBeDefined();
+    const selected: HarnessSelection = {
+      harnessId: available!.harnessId,
+      providerId: available!.providerId,
+      modelId: available!.modelId,
+    };
+
+    const handle = await adapter.create({
+      context: CONTEXT,
+      conversationId: 'conversation-1',
+      selection: selected,
+    });
+
+    const bogus: HarnessSelection = {
+      harnessId: 'fake',
+      providerId: 'ghost-provider',
+      modelId: 'ghost-model',
+    };
+
+    let error: unknown;
+    try {
+      await handle.setModel(bogus);
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessOperationError);
+    const dto = (error as HarnessOperationError).dto;
+    expect(dto.code).toBe('selection_invalid');
+    // The DTO echoes the exact requested tuple, unchanged.
+    expect(dto.selection).toEqual(bogus);
+    // No substitution to the first catalog row.
+    expect(dto.selection).not.toEqual({
+      harnessId: firstRow.harnessId,
+      providerId: firstRow.providerId,
+      modelId: firstRow.modelId,
+    });
+    // The active selection is untouched by the rejected request.
+    expect((await handle.snapshot()).selection).toEqual(selected);
+  });
+
+  it('reports model_unavailable with the unchanged tuple for a known but unavailable model', async () => {
+    const adapter = new FakeHarnessAdapter({ id: 'fake' });
+    const catalog = await adapter.catalog(CONTEXT);
+    const unavailable = catalog.models.find((entry) => entry.availability === 'unavailable');
+    const available = catalog.models.find((entry) => entry.availability === 'available');
+    expect(unavailable).toBeDefined();
+    expect(available).toBeDefined();
+
+    const startingSelection: HarnessSelection = {
+      harnessId: available!.harnessId,
+      providerId: available!.providerId,
+      modelId: available!.modelId,
+    };
+    const handle = await adapter.create({
+      context: CONTEXT,
+      conversationId: 'conversation-2',
+      selection: startingSelection,
+    });
+
+    const requested: HarnessSelection = {
+      harnessId: unavailable!.harnessId,
+      providerId: unavailable!.providerId,
+      modelId: unavailable!.modelId,
+    };
+
+    let error: unknown;
+    try {
+      await handle.setModel(requested);
+    } catch (caught) {
+      error = caught;
+    }
+
+    expect(error).toBeInstanceOf(HarnessOperationError);
+    const dto = (error as HarnessOperationError).dto;
+    expect(dto.code).toBe('model_unavailable');
+    expect(dto.selection).toEqual(requested);
+    expect((await handle.snapshot()).selection).toEqual(startingSelection);
+  });
+});

apps/gateway/src/harness/testing/fake-harness.adapter.ts

@@ -0,0 +1,248 @@
+import type {
+  AttachClient,
+  CreateHarnessSession,
+  HarnessAdapter,
+  HarnessActorContext,
+  HarnessCapability,
+  HarnessCatalog,
+  HarnessCatalogEntry,
+  HarnessCloseReason,
+  HarnessDescriptor,
+  HarnessEvent,
+  HarnessInteractionResponse,
+  HarnessPrompt,
+  HarnessPromptReceipt,
+  HarnessSelection,
+  HarnessSessionHandle,
+  HarnessSessionSnapshot,
+  HarnessSessionState,
+  ResumeHarnessSession,
+} from '@mosaicstack/types';
+import { HARNESS_CAPABILITIES } from '@mosaicstack/types';
+import { operationError } from '../harness.registry.js';
+
+export interface FakeHarnessAdapterOptions {
+  readonly id: string;
+  readonly capabilities?: readonly HarnessCapability[];
+  readonly catalog?: readonly HarnessCatalogEntry[];
+}
+
+const FAKE_PROVIDER = 'fake-openai';
+
+function defaultCatalog(harnessId: string): readonly HarnessCatalogEntry[] {
+  return [
+    {
+      harnessId,
+      providerId: FAKE_PROVIDER,
+      modelId: 'fake-mini',
+      displayName: 'Fake Mini',
+      reasoningCapability: false,
+      inputTypes: ['text'],
+      authState: 'ready',
+      availability: 'available',
+    },
+    {
+      harnessId,
+      providerId: FAKE_PROVIDER,
+      modelId: 'fake-pro',
+      displayName: 'Fake Pro',
+      reasoningCapability: true,
+      inputTypes: ['text', 'image'],
+      authState: 'ready',
+      availability: 'available',
+    },
+    {
+      harnessId,
+      providerId: FAKE_PROVIDER,
+      modelId: 'fake-legacy',
+      displayName: 'Fake Legacy',
+      reasoningCapability: false,
+      inputTypes: ['text'],
+      authState: 'unavailable',
+      availability: 'unavailable',
+    },
+  ];
+}
+
+function matches(entry: HarnessCatalogEntry, selection: HarnessSelection): boolean {
+  return (
+    entry.harnessId === selection.harnessId &&
+    entry.providerId === selection.providerId &&
+    entry.modelId === selection.modelId
+  );
+}
+
+/**
+ * In-memory harness session handle used by the fake adapter and by the shared
+ * conformance suite. It enforces the two invariants the real adapters must also
+ * honor: model selection is validated against the catalog and is NEVER
+ * substituted, and unsupported capabilities fail with a typed error.
+ */
+export class FakeHarnessSessionHandle implements HarnessSessionHandle {
+  private state: HarnessSessionState = 'idle';
+  private processId: string | undefined;
+  private readonly attachedClientIds = new Set<string>();
+  private readonly listeners = new Set<(event: HarnessEvent) => void>();
+
+  constructor(
+    private readonly conversationId: string,
+    private readonly nativeSessionId: string,
+    private readonly seatId: string,
+    private selection: HarnessSelection,
+    private readonly correlationId: string,
+    private readonly capabilities: readonly HarnessCapability[],
+    private readonly catalog: readonly HarnessCatalogEntry[],
+  ) {
+    this.processId = `process-${nativeSessionId}`;
+  }
+
+  async snapshot(): Promise<HarnessSessionSnapshot> {
+    return {
+      conversationId: this.conversationId,
+      nativeSessionId: this.nativeSessionId,
+      processId: this.processId,
+      seatId: this.seatId,
+      selection: this.selection,
+      state: this.state,
+      attachedClientIds: [...this.attachedClientIds],
+    };
+  }
+
+  async attach(input: AttachClient): Promise<void> {
+    this.attachedClientIds.add(input.clientId);
+  }
+
+  async detach(clientId: string): Promise<void> {
+    // Removes the browser attachment only; the process and native session persist.
+    this.attachedClientIds.delete(clientId);
+  }
+
+  async prompt(input: HarnessPrompt & { idempotencyKey: string }): Promise<HarnessPromptReceipt> {
+    return {
+      conversationId: this.conversationId,
+      turnId: input.turnId,
+      correlationId: input.correlationId,
+      state: 'accepted',
+      selection: this.selection,
+    };
+  }
+
+  async setModel(selection: HarnessSelection): Promise<HarnessSelection> {
+    const entry = this.catalog.find((candidate) => matches(candidate, selection));
+    if (!entry) {
+      // No fallback to the first catalog row: reject with the requested tuple, unchanged.
+      throw operationError(
+        'selection_invalid',
+        'The requested harness/provider/model tuple is not in the catalog.',
+        selection,
+        this.correlationId,
+      );
+    }
+    if (entry.availability === 'unavailable') {
+      throw operationError(
+        'model_unavailable',
+        'The requested model is currently unavailable.',
+        selection,
+        this.correlationId,
+        true,
+      );
+    }
+    this.selection = selection;
+    return this.selection;
+  }
+
+  async abort(_turnId: string): Promise<void> {
+    // No active turn machinery in the fake; abort is a no-op acknowledgement.
+  }
+
+  async respondInteraction(_input: HarnessInteractionResponse): Promise<void> {
+    if (!this.capabilities.includes('extensionUi')) {
+      throw operationError(
+        'interaction_unsupported',
+        'This harness does not support interactive responses.',
+        this.selection,
+        this.correlationId,
+      );
+    }
+  }
+
+  events(listener: (event: HarnessEvent) => void): () => void {
+    this.listeners.add(listener);
+    return () => {
+      this.listeners.delete(listener);
+    };
+  }
+
+  async evictProcess(_reason: HarnessCloseReason): Promise<void> {
+    // Stop the process but keep the resumable native session.
+    this.processId = undefined;
+    this.state = 'evicted';
+  }
+
+  async endSession(_reason: HarnessCloseReason): Promise<void> {
+    // Destructively end the native session.
+    this.processId = undefined;
+    this.state = 'ended';
+  }
+}
+
+/**
+ * Minimal in-memory {@link HarnessAdapter} for Slice Zero. It mints a fresh
+ * native session id on `create` and binds the supplied one on `resume`, so the
+ * two paths are observably distinct.
+ */
+export class FakeHarnessAdapter implements HarnessAdapter {
+  readonly id: string;
+  private readonly capabilities: readonly HarnessCapability[];
+  private readonly catalogEntries: readonly HarnessCatalogEntry[];
+  private createdCount = 0;
+
+  constructor(options: FakeHarnessAdapterOptions) {
+    this.id = options.id;
+    this.capabilities = options.capabilities ?? [...HARNESS_CAPABILITIES];
+    this.catalogEntries = options.catalog ?? defaultCatalog(options.id);
+  }
+
+  async describe(_context: HarnessActorContext): Promise<HarnessDescriptor> {
+    return {
+      id: this.id,
+      displayName: `Fake harness (${this.id})`,
+      capabilities: this.capabilities,
+    };
+  }
+
+  async catalog(_context: HarnessActorContext): Promise<HarnessCatalog> {
+    return {
+      harnessId: this.id,
+      version: '1.0.0',
+      fingerprint: `fake-${this.id}-${this.catalogEntries.length}`,
+      models: this.catalogEntries,
+    };
+  }
+
+  async create(input: CreateHarnessSession): Promise<HarnessSessionHandle> {
+    this.createdCount += 1;
+    const nativeSessionId = `native-${input.conversationId}-${this.createdCount}`;
+    return new FakeHarnessSessionHandle(
+      input.conversationId,
+      nativeSessionId,
+      input.context.seatId,
+      input.selection,
+      input.context.correlationId,
+      this.capabilities,
+      this.catalogEntries,
+    );
+  }
+
+  async resume(input: ResumeHarnessSession): Promise<HarnessSessionHandle> {
+    return new FakeHarnessSessionHandle(
+      input.conversationId,
+      input.nativeSessionId,
+      input.context.seatId,
+      input.selection,
+      input.context.correlationId,
+      this.capabilities,
+      this.catalogEntries,
+    );
+  }
+}

apps/gateway/src/harness/testing/harness-adapter.contract.ts

@@ -0,0 +1,157 @@
+import { describe, expect, it } from 'vitest';
+import type {
+  HarnessActorContext,
+  HarnessAdapter,
+  HarnessCatalogEntry,
+  HarnessSelection,
+} from '@mosaicstack/types';
+import { HarnessOperationError } from '../harness.registry.js';
+
+const CONTEXT: HarnessActorContext = {
+  actorId: 'contract-actor',
+  tenantId: 'contract-tenant',
+  seatId: 'contract-seat',
+  correlationId: 'contract-correlation',
+};
+
+function toSelection(entry: HarnessCatalogEntry): HarnessSelection {
+  return { harnessId: entry.harnessId, providerId: entry.providerId, modelId: entry.modelId };
+}
+
+function pickAvailable(models: readonly HarnessCatalogEntry[]): HarnessCatalogEntry {
+  const entry = models.find((candidate) => candidate.availability === 'available') ?? models[0];
+  if (!entry) {
+    throw new Error('contract fixture requires at least one catalog model');
+  }
+  return entry;
+}
+
+async function captureError(run: () => Promise<unknown>): Promise<unknown> {
+  try {
+    await run();
+    return undefined;
+  } catch (caught) {
+    return caught;
+  }
+}
+
+/**
+ * Shared conformance suite every {@link HarnessAdapter} must pass. Slice Zero
+ * runs it against the fake adapter; Task 13 re-runs the identical suite against
+ * the native Pi adapter so both share one behavioral contract.
+ */
+export function runHarnessAdapterContract(
+  label: string,
+  createAdapter: () => HarnessAdapter,
+): void {
+  describe(`harness adapter contract: ${label}`, () => {
+    it('mints a fresh native session on create and binds the supplied one on resume', async () => {
+      const adapter = createAdapter();
+      const catalog = await adapter.catalog(CONTEXT);
+      const selection = toSelection(pickAvailable(catalog.models));
+
+      const created = await (
+        await adapter.create({ context: CONTEXT, conversationId: 'conv-create', selection })
+      ).snapshot();
+      const resumed = await (
+        await adapter.resume({
+          context: CONTEXT,
+          conversationId: 'conv-resume',
+          nativeSessionId: 'native-supplied-1',
+          selection,
+        })
+      ).snapshot();
+
+      expect(created.nativeSessionId).toBeTruthy();
+      expect(resumed.nativeSessionId).toBe('native-supplied-1');
+      expect(created.nativeSessionId).not.toBe(resumed.nativeSessionId);
+      expect(created.seatId).toBe(CONTEXT.seatId);
+    });
+
+    it('gives detach, evict, and end distinct effects (not aliases)', async () => {
+      const adapter = createAdapter();
+      const catalog = await adapter.catalog(CONTEXT);
+      const selection = toSelection(pickAvailable(catalog.models));
+      const handle = await adapter.create({
+        context: CONTEXT,
+        conversationId: 'conv-lifecycle',
+        selection,
+      });
+
+      await handle.attach({ clientId: 'browser-1' });
+      await handle.detach('browser-1');
+      const afterDetach = await handle.snapshot();
+      expect(afterDetach.attachedClientIds).toEqual([]);
+      expect(afterDetach.state).not.toBe('evicted');
+      expect(afterDetach.state).not.toBe('ended');
+
+      await handle.evictProcess('idle_timeout');
+      const afterEvict = await handle.snapshot();
+      expect(afterEvict.state).toBe('evicted');
+      // The native session survives eviction (resumable); the process does not.
+      expect(afterEvict.nativeSessionId).toBe(afterDetach.nativeSessionId);
+      expect(afterEvict.processId).toBeUndefined();
+
+      await handle.endSession('session_ended');
+      const afterEnd = await handle.snapshot();
+      expect(afterEnd.state).toBe('ended');
+      // End is not an alias of evict.
+      expect(afterEnd.state).not.toBe(afterEvict.state);
+    });
+
+    it('never substitutes the first catalog row for an unknown selection', async () => {
+      const adapter = createAdapter();
+      const catalog = await adapter.catalog(CONTEXT);
+      const firstRow = catalog.models[0];
+      if (!firstRow) {
+        throw new Error('contract fixture requires a catalog model');
+      }
+      const start = toSelection(pickAvailable(catalog.models));
+      const handle = await adapter.create({
+        context: CONTEXT,
+        conversationId: 'conv-nosub',
+        selection: start,
+      });
+
+      const bogus: HarnessSelection = {
+        harnessId: adapter.id,
+        providerId: 'contract-ghost-provider',
+        modelId: 'contract-ghost-model',
+      };
+      const error = await captureError(() => handle.setModel(bogus));
+
+      expect(error).toBeInstanceOf(HarnessOperationError);
+      const dto = (error as HarnessOperationError).dto;
+      expect(dto.code).toBe('selection_invalid');
+      expect(dto.selection).toEqual(bogus);
+      expect(dto.selection).not.toEqual(toSelection(firstRow));
+      expect((await handle.snapshot()).selection).toEqual(start);
+    });
+
+    it('validates capability-gated interactions with a typed error, not a silent no-op', async () => {
+      const adapter = createAdapter();
+      const descriptor = await adapter.describe(CONTEXT);
+      const catalog = await adapter.catalog(CONTEXT);
+      const selection = toSelection(pickAvailable(catalog.models));
+      const handle = await adapter.create({
+        context: CONTEXT,
+        conversationId: 'conv-interaction',
+        selection,
+      });
+
+      const response = {
+        requestId: 'interaction-1',
+        type: 'confirm',
+        accepted: true,
+      } as const;
+
+      if (descriptor.capabilities.includes('extensionUi')) {
+        await expect(handle.respondInteraction(response)).resolves.toBeUndefined();
+      } else {
+        const error = await captureError(() => handle.respondInteraction(response));
+        expect(error).toBeInstanceOf(HarnessOperationError);
+        expect((error as HarnessOperationError).dto.code).toBe('interaction_unsupported');
+      }
+    });
+  });
+}

apps/gateway/src/main.spec.ts

@@ -0,0 +1,164 @@
+import 'reflect-metadata';
+import { mkdir, mkdtemp, rm, writeFile } from 'node:fs/promises';
+import * as nodeOs from 'node:os';
+import { dirname, isAbsolute, join, relative, resolve } from 'node:path';
+import * as nodeUrl from 'node:url';
+import type { MosaicConfig } from '@mosaicstack/config';
+import type * as MosaicStorage from '@mosaicstack/storage';
+import { describe, expect, it, vi, type MockInstance } from 'vitest';
+
+// Each case uses vi.resetModules() and re-imports the full gateway graph for distinct ambient FS/env; CI needs headroom, while this still guards genuine hangs.
+const MODULE_IMPORT_TIMEOUT_MS = 120_000;
+
+function snapshotProcessEnv(): Record<string, string | undefined> {
+  return { ...process.env };
+}
+
+function restoreProcessEnv(snapshot: Record<string, string | undefined>): void {
+  for (const key of Object.keys(process.env)) {
+    if (!(key in snapshot)) {
+      delete process.env[key];
+    }
+  }
+
+  for (const [key, value] of Object.entries(snapshot)) {
+    if (value === undefined) {
+      delete process.env[key];
+      continue;
+    }
+
+    process.env[key] = value;
+  }
+}
+
+function expectPathUnderTempRoot(path: string, tempRoot: string): void {
+  const relativePath = relative(tempRoot, path);
+  expect(relativePath === '' || (!relativePath.startsWith('..') && !isAbsolute(relativePath))).toBe(
+    true,
+  );
+}
+
+async function writeFixture(path: string, contents: string, tempRoot: string): Promise<void> {
+  expectPathUnderTempRoot(path, tempRoot);
+  await mkdir(dirname(path), { recursive: true });
+  await writeFile(path, contents, 'utf8');
+}
+
+interface BootstrapPreflightResult {
+  capturedConfig: MosaicConfig | undefined;
+}
+
+async function runBootstrapPreflight(
+  anchoredConfigContents: string,
+  ambientConfigContents: string,
+): Promise<BootstrapPreflightResult> {
+  const originalEnv = snapshotProcessEnv();
+  const tempRoot = await mkdtemp(join(nodeOs.tmpdir(), 'mosaic-gateway-main-preflight-'));
+  let cwdSpy: ReturnType<typeof vi.spyOn> | undefined;
+  let exitSpy: MockInstance<typeof process.exit> | undefined;
+  let consoleInfoSpy: ReturnType<typeof vi.spyOn> | undefined;
+  let capturedConfig: MosaicConfig | undefined;
+
+  try {
+    const anchor = join(tempRoot, 'anchored', 'apps', 'gateway', 'src');
+    const homePath = join(tempRoot, 'home');
+    const cwdPath = join(tempRoot, 'ambient', 'cwd');
+    const monorepoRootConfigPath = resolve(anchor, '../../..', 'mosaic.config.json');
+
+    await mkdir(anchor, { recursive: true });
+    await mkdir(cwdPath, { recursive: true });
+
+    await writeFixture(monorepoRootConfigPath, anchoredConfigContents, tempRoot);
+    await writeFixture(join(cwdPath, 'mosaic.config.json'), ambientConfigContents, tempRoot);
+
+    process.env['HOME'] = homePath;
+    process.env['BETTER_AUTH_SECRET'] = 'fixture-secret';
+    delete process.env['MOSAIC_STORAGE_TIER'];
+    delete process.env['DATABASE_URL'];
+    delete process.env['VALKEY_URL'];
+
+    consoleInfoSpy = vi.spyOn(console, 'info').mockImplementation((): void => undefined);
+    const exitMock = vi.fn<typeof process.exit>();
+    exitSpy = vi.spyOn(process, 'exit').mockImplementation(exitMock);
+
+    vi.resetModules();
+    vi.doMock('node:os', () => ({ ...nodeOs, homedir: (): string => homePath }));
+    vi.doMock('node:url', () => ({
+      ...nodeUrl,
+      fileURLToPath: (url: string | URL): string => {
+        const actualPath = nodeUrl.fileURLToPath(url);
+        if (
+          actualPath.endsWith('/apps/gateway/src/env.ts') ||
+          actualPath.endsWith('/apps/gateway/src/env.js')
+        ) {
+          return join(anchor, 'env.ts');
+        }
+        return actualPath;
+      },
+    }));
+    cwdSpy = vi.spyOn(process, 'cwd').mockReturnValue(cwdPath);
+    vi.doMock('./tracing.js', () => ({}));
+
+    const preflightSentinel = new Error('preflight-capture-sentinel');
+    vi.doMock('@mosaicstack/storage', async () => {
+      const actual = await vi.importActual<typeof MosaicStorage>('@mosaicstack/storage');
+      return {
+        ...actual,
+        detectAndAssertTier: vi.fn((config: MosaicConfig): Promise<void> => {
+          capturedConfig = config;
+          throw preflightSentinel;
+        }),
+      };
+    });
+
+    await import('./main.js');
+    await vi.waitFor((): void => {
+      expect(exitSpy).toHaveBeenCalled();
+    });
+
+    return { capturedConfig };
+  } finally {
+    cwdSpy?.mockRestore();
+    exitSpy?.mockRestore();
+    consoleInfoSpy?.mockRestore();
+    vi.doUnmock('@mosaicstack/storage');
+    vi.doUnmock('./tracing.js');
+    vi.doUnmock('node:url');
+    vi.doUnmock('node:os');
+    vi.resetModules();
+    restoreProcessEnv(originalEnv);
+    await rm(tempRoot, { recursive: true, force: true });
+  }
+}
+
+describe('main bootstrap preflight config anchoring', (): void => {
+  it(
+    'passes the anchored monorepo-root config to detectAndAssertTier, not an ambient cwd config',
+    async (): Promise<void> => {
+      const anchoredConfig = JSON.stringify({
+        tier: 'local',
+        storage: { type: 'pglite', dataDir: '.mosaic/storage-pglite' },
+        queue: { type: 'local', dataDir: '.mosaic/queue' },
+        memory: { type: 'keyword' },
+      });
+      const ambientConfig = JSON.stringify({
+        tier: 'federated',
+        storage: {
+          type: 'postgres',
+          url: 'postgresql://ambient-attacker.invalid/mosaic',
+          enableVector: true,
+        },
+        queue: { type: 'bullmq' },
+        memory: { type: 'pgvector' },
+      });
+
+      const { capturedConfig } = await runBootstrapPreflight(anchoredConfig, ambientConfig);
+
+      expect(capturedConfig?.tier).toBe('local');
+      expect(capturedConfig?.storage).not.toEqual(
+        expect.objectContaining({ url: 'postgresql://ambient-attacker.invalid/mosaic' }),
+      );
+    },
+    MODULE_IMPORT_TIMEOUT_MS,
+  );
+});

apps/gateway/src/main.ts

@@ -1,18 +1,5 @@
 #!/usr/bin/env node
-import { config } from 'dotenv';
+import './env.js';
-import { existsSync } from 'node:fs';
-import { resolve, join } from 'node:path';
-import { homedir } from 'node:os';
-
-// Load .env from daemon config dir (global install / daemon mode).
-// Loaded first so monorepo .env can override for local dev.
-const daemonEnv = join(homedir(), '.config', 'mosaic', 'gateway', '.env');
-if (existsSync(daemonEnv)) config({ path: daemonEnv });
-
-// Load .env from monorepo root (cwd is apps/gateway when run via pnpm filter)
-config({ path: resolve(process.cwd(), '../../.env') });
-config(); // Also load apps/gateway/.env if present (overrides)
-
 import './tracing.js';
 import 'reflect-metadata';
 import { NestFactory } from '@nestjs/core';
@@ -26,6 +13,7 @@ import { mountAuthHandler } from './auth/auth.controller.js';
 import { mountMcpHandler } from './mcp/mcp.controller.js';
 import { McpService } from './mcp/mcp.service.js';
 import { detectAndAssertTier, TierDetectionError } from '@mosaicstack/storage';
+import { resolveGatewayConfigPath } from './env.js';
 
 async function bootstrap(): Promise<void> {
   const logger = new Logger('Bootstrap');
@@ -37,7 +25,7 @@ async function bootstrap(): Promise<void> {
   // Pre-flight: assert all external services required by the configured tier
   // are reachable. Runs before NestFactory.create() so failures are visible
   // immediately with actionable remediation hints.
-  const mosaicConfig = loadConfig();
+  const mosaicConfig = loadConfig(resolveGatewayConfigPath());
   try {
     await detectAndAssertTier(mosaicConfig);
   } catch (err) {

apps/gateway/src/mcp-client/mcp-client.service.spec.ts

@@ -0,0 +1,44 @@
+import { Logger } from '@nestjs/common';
+import { Client } from '@modelcontextprotocol/sdk/client/index.js';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+import { McpClientService } from './mcp-client.service.js';
+
+const MCP_LEAK_MARKER = 'MCP_LEAK_MARKER /srv/secret';
+
+describe('McpClientService — failed connect error sanitization', () => {
+  const originalMcpServers = process.env['MCP_SERVERS'];
+
+  beforeEach(() => {
+    process.env['MCP_SERVERS'] = JSON.stringify([
+      { name: 'leaky-server', url: 'http://localhost:9999/mcp' },
+    ]);
+  });
+
+  afterEach(() => {
+    vi.restoreAllMocks();
+    if (originalMcpServers === undefined) {
+      delete process.env['MCP_SERVERS'];
+    } else {
+      process.env['MCP_SERVERS'] = originalMcpServers;
+    }
+  });
+
+  it('stores a generic serverEntry.error while logging the raw exception server-side', async () => {
+    vi.spyOn(Client.prototype, 'connect').mockRejectedValue(new Error(MCP_LEAK_MARKER));
+    const errorSpy = vi.spyOn(Logger.prototype, 'error').mockImplementation(() => undefined);
+
+    const service = new McpClientService();
+    await service.onModuleInit();
+
+    const statuses = service.getServerStatuses();
+    expect(statuses).toHaveLength(1);
+    expect(statuses[0]?.connected).toBe(false);
+    expect(statuses[0]?.error).toBe('Connection failed (see server logs).');
+    expect(statuses[0]?.error).not.toContain(MCP_LEAK_MARKER);
+
+    const loggedRawMarker = errorSpy.mock.calls.some((call) =>
+      call.some((arg) => typeof arg === 'string' && arg.includes(MCP_LEAK_MARKER)),
+    );
+    expect(loggedRawMarker).toBe(true);
+  });
+});

apps/gateway/src/mcp-client/mcp-client.service.ts

@@ -189,7 +189,7 @@ export class McpClientService implements OnModuleInit, OnModuleDestroy {
       );
     } catch (err) {
       const message = err instanceof Error ? err.message : String(err);
-      serverEntry.error = message;
+      serverEntry.error = 'Connection failed (see server logs).';
       serverEntry.connected = false;
       this.logger.error(`Failed to connect to MCP server "${config.name}": ${message}`);
     }

apps/gateway/src/plugin/discord-ingress.security.spec.ts

@@ -12,6 +12,10 @@ import { RuntimeProviderService } from '../agent/runtime-provider-registry.servi
 import { ChatGateway } from '../chat/chat.gateway.js';
 import { CommandAuthorizationService } from '../commands/command-authorization.service.js';
 import { validateDiscordServiceToken } from '../chat/chat.gateway-auth.js';
+import { ChatRuntimeRouter } from '../chat/chat-runtime-router.js';
+import { EmbeddedChatRuntime } from '../chat/embedded-chat.runtime.js';
+import { HarnessChatRuntime } from '../chat/harness-chat.runtime.js';
+import { HarnessRegistry } from '../harness/harness.registry.js';
 import { DiscordReplayProtector } from './discord-replay-protector.js';
 
 const SERVICE_TOKEN = 'test-service-token';
@@ -25,6 +29,7 @@ const ENV_KEYS = [
   'DISCORD_ALLOWED_USER_IDS',
   'MOSAIC_AGENT_NAME',
   'MOSAIC_AGENT_CONFIG_ID',
+  'CHAT_HARNESS_RUNTIME',
 ] as const;
 const savedEnv = new Map<string, string | undefined>();
 
@@ -150,6 +155,57 @@ function createPayload(overrides: Partial<DiscordIngressPayload> = {}): DiscordI
   };
 }
 
+/**
+ * Task 5 fence (C): the Discord SEND path runs through the exclusive {@link ChatRuntimeRouter},
+ * constructed here in `pi-rpc` mode with a fully-resolved runtime (`active` = harness). A verified
+ * Discord *service* turn must nonetheless execute on the {@link EmbeddedChatRuntime} — never the
+ * harness, never the routing engine — per the Q1/Q2 adjudication: the router owns a dedicated
+ * verified-ingress dispatch that delegates to embedded regardless of mode, with zero harness
+ * fallback. The gateway is given the router in the former direct-`AgentService` constructor slot.
+ *
+ * RED today: production still reads that slot as a bare `AgentService`, so `this.agentService`
+ * resolves to the router, `getSession(...)` is not a function, the send path throws and is caught
+ * (an `error` is emitted and the handler returns) BEFORE it ever reaches the embedded runtime. The
+ * failure is behavioural wiring — collection, DI, and `onModuleInit` all succeed. GREEN re-routes
+ * the verified Discord dispatch through the router into the embedded runtime, satisfying the
+ * preserved create/prompt assertions without weakening any control. `harnessConversations.append`
+ * proves the harness path is never touched even though the pi-rpc router resolved it as `active`.
+ *
+ * Correction #4 is proved behaviourally, not by naming an accessor: the verified-ingress dispatch
+ * is reachable only from the fully-verified `discordService` branch (the create/prompt tests below)
+ * and never from a browser-emittable socket event (the browser-forgery refusal test).
+ */
+function readyPiRpcRegistry(): HarnessRegistry {
+  const registry = new HarnessRegistry();
+  // A registered 'pi' adapter + an available (non-sentinel) conversation service let the pi-rpc
+  // router resolve `active` = harness instead of failing closed at init, so these tests model the
+  // real hostile condition — the harness runtime IS live — rather than a degraded router.
+  registry.register({ id: 'pi' } as never);
+  return registry;
+}
+
+function piRpcRouterFronting(
+  agentService: unknown,
+  harnessConversations: { append: ReturnType<typeof vi.fn> },
+): ChatRuntimeRouter {
+  const routerConversationServiceTripwire = {
+    append: () => {
+      throw new Error('router conversation service must not be resolved on the Discord path');
+    },
+  };
+  const embedded = new EmbeddedChatRuntime(agentService as never);
+  const harness = new HarnessChatRuntime(harnessConversations as never);
+  const router = new ChatRuntimeRouter(
+    readyPiRpcRegistry(),
+    routerConversationServiceTripwire as never,
+    embedded,
+    harness,
+    'pi-rpc',
+  );
+  router.onModuleInit();
+  return router;
+}
+
 describe('Discord ingress security', () => {
   it('keeps legacy role-only bindings valid while withholding privileged actor identity', () => {
     const [binding] = parseDiscordInteractionBindings(
@@ -433,6 +489,7 @@ describe('Discord ingress security', () => {
 
   it("selects each binding's trusted logical-agent config when creating Discord sessions", async () => {
     configureDiscordEnv();
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
     process.env['DISCORD_ALLOWED_CHANNEL_IDS'] = 'channel-001,channel-002';
     process.env['DISCORD_INTERACTION_BINDINGS'] = JSON.stringify([
       {
@@ -489,8 +546,9 @@ describe('Discord ingress security', () => {
       },
     };
     const routingEngine = { resolve: vi.fn() };
+    const harnessConversations = { append: vi.fn() };
     const gateway = new ChatGateway(
-      agentService as never,
+      piRpcRouterFronting(agentService, harnessConversations) as never,
       {} as never,
       brain as never,
       {} as never,
@@ -531,6 +589,575 @@ describe('Discord ingress security', () => {
       expect.objectContaining({ agentConfigId: 'agent-config-orion' }),
     );
     expect(routingEngine.resolve).not.toHaveBeenCalled();
+    // Even though the pi-rpc router resolved the harness as `active`, verified Discord ingress must
+    // never touch it — the create path stays on the embedded runtime.
+    expect(harnessConversations.append).not.toHaveBeenCalled();
+  });
+
+  it('dispatches a verified Discord SEND once and drops a byte-identical replay with zero additional dispatch/persist/ack (Task 5 G4)', async () => {
+    configureDiscordEnv();
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
+    process.env['DISCORD_ALLOWED_CHANNEL_IDS'] = 'channel-001';
+    process.env['DISCORD_INTERACTION_BINDINGS'] = JSON.stringify([
+      {
+        instanceId: 'Nova',
+        agentConfigId: 'agent-config-nova',
+        guildId: 'guild-001',
+        channelId: 'channel-001',
+        pairedUsers: {
+          'user-001': { role: 'operator', mosaicUserId: 'mosaic-operator-001' },
+        },
+      },
+    ]);
+    const session = {
+      provider: 'configured-provider',
+      modelId: 'configured-model',
+      agentConfigId: 'agent-config-nova',
+      agentName: 'Nova',
+      piSession: {
+        thinkingLevel: 'medium',
+        getAvailableThinkingLevels: (): string[] => ['medium'],
+      },
+    };
+    const createSession = vi.fn().mockResolvedValue(session);
+    const prompt = vi.fn().mockResolvedValue(undefined);
+    const agentService = {
+      getSession: vi.fn().mockReturnValue(undefined),
+      createSession,
+      recordMessage: vi.fn(),
+      onEvent: vi.fn().mockReturnValue((): void => undefined),
+      addChannel: vi.fn(),
+      removeChannel: vi.fn(),
+      prompt,
+    };
+    const addMessage = vi.fn().mockResolvedValue({ id: 'discord-persisted-message' });
+    const brain = {
+      agents: { findById: vi.fn((id: string) => Promise.resolve({ id, name: 'Nova' })) },
+      conversations: {
+        findById: vi.fn().mockResolvedValue({ id: 'Nova:discord:channel-001' }),
+        findMessages: vi.fn().mockResolvedValue([]),
+        create: vi.fn().mockResolvedValue(undefined),
+        update: vi.fn().mockResolvedValue(undefined),
+        addMessage,
+      },
+    };
+    const harnessConversations = { append: vi.fn() };
+    const gateway = new ChatGateway(
+      piRpcRouterFronting(agentService, harnessConversations) as never,
+      {} as never,
+      brain as never,
+      {} as never,
+      {} as never,
+      { resolve: vi.fn() } as never,
+    );
+    const client = {
+      id: 'discord-client-replay',
+      data: { discordService: true },
+      emit: vi.fn(),
+    };
+    const ackCount = (): number =>
+      client.emit.mock.calls.filter((call) => call[0] === 'message:ack').length;
+
+    // One fully-valid signed envelope; the replay reuses the SAME object (same messageId).
+    const envelope = ingressEnvelope('verified once', 'discord-replay-001', {
+      conversationId: 'Nova:discord:channel-001',
+    });
+
+    // First delivery: the verified-Discord SEND runs the full embedded dispatch exactly once.
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(1);
+    expect(prompt).toHaveBeenCalledTimes(1);
+    expect(addMessage).toHaveBeenCalledTimes(1);
+    expect(ackCount()).toBe(1);
+
+    // Byte-identical replay: the messageId is already claimed, so resolveDiscordIngress returns
+    // null and the SEND handler bails before dispatch/persist/ack. Every effect stays at exactly one.
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(1);
+    expect(prompt).toHaveBeenCalledTimes(1);
+    expect(addMessage).toHaveBeenCalledTimes(1);
+    expect(ackCount()).toBe(1);
+    // The harness runtime is never touched on either delivery.
+    expect(harnessConversations.append).not.toHaveBeenCalled();
+  });
+
+  it('a verified SEND that fails the configured service identity consumes no replay claim, so a corrected byte-identical retry dispatches/persists/acks exactly once and a later duplicate stays fail-closed (Task 5 item 4 — claim ordering)', async () => {
+    configureDiscordEnv();
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
+    process.env['DISCORD_ALLOWED_CHANNEL_IDS'] = 'channel-001';
+    process.env['DISCORD_INTERACTION_BINDINGS'] = JSON.stringify([
+      {
+        instanceId: 'Nova',
+        agentConfigId: 'agent-config-nova',
+        guildId: 'guild-001',
+        channelId: 'channel-001',
+        pairedUsers: {
+          'user-001': { role: 'operator', mosaicUserId: 'mosaic-operator-001' },
+        },
+      },
+    ]);
+    const session = {
+      provider: 'configured-provider',
+      modelId: 'configured-model',
+      agentConfigId: 'agent-config-nova',
+      agentName: 'Nova',
+      piSession: {
+        thinkingLevel: 'medium',
+        getAvailableThinkingLevels: (): string[] => ['medium'],
+      },
+    };
+    const createSession = vi.fn().mockResolvedValue(session);
+    const prompt = vi.fn().mockResolvedValue(undefined);
+    const agentService = {
+      getSession: vi.fn().mockReturnValue(undefined),
+      createSession,
+      recordMessage: vi.fn(),
+      onEvent: vi.fn().mockReturnValue((): void => undefined),
+      addChannel: vi.fn(),
+      removeChannel: vi.fn(),
+      prompt,
+    };
+    const addMessage = vi.fn().mockResolvedValue({ id: 'discord-persisted-message' });
+    const brain = {
+      agents: { findById: vi.fn((id: string) => Promise.resolve({ id, name: 'Nova' })) },
+      conversations: {
+        findById: vi.fn().mockResolvedValue({ id: 'Nova:discord:channel-001' }),
+        findMessages: vi.fn().mockResolvedValue([]),
+        create: vi.fn().mockResolvedValue(undefined),
+        update: vi.fn().mockResolvedValue(undefined),
+        addMessage,
+      },
+    };
+    const harnessConversations = { append: vi.fn() };
+    const gateway = new ChatGateway(
+      piRpcRouterFronting(agentService, harnessConversations) as never,
+      {} as never,
+      brain as never,
+      {} as never,
+      {} as never,
+      { resolve: vi.fn() } as never,
+    );
+    const client = {
+      id: 'discord-client-claim-ordering',
+      data: { discordService: true },
+      emit: vi.fn(),
+    };
+    const ackCount = (): number =>
+      client.emit.mock.calls.filter((call) => call[0] === 'message:ack').length;
+
+    // A single fully-valid signed envelope, reused byte-for-byte across all three deliveries.
+    const envelope = ingressEnvelope('verified once with late identity', 'discord-order-001', {
+      conversationId: 'Nova:discord:channel-001',
+    });
+
+    // (1) Configured service identity is MISSING. The envelope is validly signed and passes the
+    //     binding + route checks, but the SEND must refuse at the identity gate BEFORE any claim
+    //     or effect. If the claim fires ahead of that gate, this delivery silently burns the
+    //     replay claim for `discord-order-001` even though nothing dispatched.
+    delete process.env['DISCORD_SERVICE_USER_ID'];
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(0);
+    expect(prompt).toHaveBeenCalledTimes(0);
+    expect(addMessage).toHaveBeenCalledTimes(0);
+    expect(ackCount()).toBe(0);
+
+    // (2) Identity is now configured; the operator resends the SAME envelope byte-for-byte. Because
+    //     step (1) consumed no claim, this corrected retry claims once and runs the full embedded
+    //     dispatch exactly once. (Under the pre-fix ordering the claim was already spent in step (1),
+    //     so this retry is dropped as a replay and never dispatches — the RED this test drives.)
+    process.env['DISCORD_SERVICE_USER_ID'] = 'discord-service';
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(1);
+    expect(prompt).toHaveBeenCalledTimes(1);
+    expect(addMessage).toHaveBeenCalledTimes(1);
+    expect(ackCount()).toBe(1);
+
+    // (3) A genuine duplicate after a committed turn stays fail-closed: the claim taken in step (2)
+    //     blocks it, so every effect remains at exactly one.
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(1);
+    expect(prompt).toHaveBeenCalledTimes(1);
+    expect(addMessage).toHaveBeenCalledTimes(1);
+    expect(ackCount()).toBe(1);
+    expect(harnessConversations.append).not.toHaveBeenCalled();
+  });
+
+  it('a verified SEND whose configured agent record fails reconciliation consumes no replay claim, so a corrected byte-identical retry dispatches/persists/acks exactly once (Task 5 finding 3)', async () => {
+    configureDiscordEnv();
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
+    process.env['DISCORD_ALLOWED_CHANNEL_IDS'] = 'channel-001';
+    process.env['DISCORD_INTERACTION_BINDINGS'] = JSON.stringify([
+      {
+        instanceId: 'Nova',
+        agentConfigId: 'agent-config-nova',
+        guildId: 'guild-001',
+        channelId: 'channel-001',
+        pairedUsers: {
+          'user-001': { role: 'operator', mosaicUserId: 'mosaic-operator-001' },
+        },
+      },
+    ]);
+    const session = {
+      provider: 'configured-provider',
+      modelId: 'configured-model',
+      agentConfigId: 'agent-config-nova',
+      agentName: 'Nova',
+      piSession: {
+        thinkingLevel: 'medium',
+        getAvailableThinkingLevels: (): string[] => ['medium'],
+      },
+    };
+    const createSession = vi.fn().mockResolvedValue(session);
+    const prompt = vi.fn().mockResolvedValue(undefined);
+    const agentService = {
+      getSession: vi.fn().mockReturnValue(undefined),
+      createSession,
+      recordMessage: vi.fn(),
+      onEvent: vi.fn().mockReturnValue((): void => undefined),
+      addChannel: vi.fn(),
+      removeChannel: vi.fn(),
+      prompt,
+    };
+    const addMessage = vi.fn().mockResolvedValue({ id: 'discord-persisted-message' });
+    // The durable agent record does not reconcile on the first delivery (its name no longer matches
+    // the verified binding's instance id), then reconciles cleanly on the corrected retry.
+    const findAgent = vi
+      .fn()
+      .mockResolvedValueOnce({ id: 'agent-config-nova', name: 'Renamed-Away' })
+      .mockResolvedValue({ id: 'agent-config-nova', name: 'Nova' });
+    const brain = {
+      agents: { findById: findAgent },
+      conversations: {
+        findById: vi.fn().mockResolvedValue({ id: 'Nova:discord:channel-001' }),
+        findMessages: vi.fn().mockResolvedValue([]),
+        create: vi.fn().mockResolvedValue(undefined),
+        update: vi.fn().mockResolvedValue(undefined),
+        addMessage,
+      },
+    };
+    const harnessConversations = { append: vi.fn() };
+    const gateway = new ChatGateway(
+      piRpcRouterFronting(agentService, harnessConversations) as never,
+      {} as never,
+      brain as never,
+      {} as never,
+      {} as never,
+      { resolve: vi.fn() } as never,
+    );
+    const client = {
+      id: 'discord-client-reconcile',
+      data: { discordService: true },
+      emit: vi.fn(),
+    };
+    const ackCount = (): number =>
+      client.emit.mock.calls.filter((call) => call[0] === 'message:ack').length;
+
+    const envelope = ingressEnvelope(
+      'verified once with stale agent record',
+      'discord-reconcile-001',
+      {
+        conversationId: 'Nova:discord:channel-001',
+      },
+    );
+
+    // (1) The configured-agent reconcile runs BEFORE the replay claim. A mismatch refuses the turn
+    //     and, crucially, consumes no claim for discord-reconcile-001 — nothing dispatches.
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(0);
+    expect(prompt).toHaveBeenCalledTimes(0);
+    expect(addMessage).toHaveBeenCalledTimes(0);
+    expect(ackCount()).toBe(0);
+
+    // (2) The record now reconciles; because step (1) took no claim, this byte-identical retry claims
+    //     once and runs the full embedded dispatch exactly once. (Pre-fix, the claim was spent ahead
+    //     of the reconcile in step (1), so this retry was dropped as a replay — the RED this drives.)
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(1);
+    expect(prompt).toHaveBeenCalledTimes(1);
+    expect(addMessage).toHaveBeenCalledTimes(1);
+    expect(ackCount()).toBe(1);
+
+    // (3) A genuine duplicate after the committed turn stays fail-closed.
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(1);
+    expect(prompt).toHaveBeenCalledTimes(1);
+    expect(addMessage).toHaveBeenCalledTimes(1);
+    expect(ackCount()).toBe(1);
+    expect(harnessConversations.append).not.toHaveBeenCalled();
+  });
+
+  it('a verified SEND refuses to reuse a same-scope embedded session minted under a different configured identity, with zero prompt/persist/ack (Task 5 finding 3)', async () => {
+    configureDiscordEnv();
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
+    process.env['DISCORD_ALLOWED_CHANNEL_IDS'] = 'channel-001';
+    process.env['DISCORD_INTERACTION_BINDINGS'] = JSON.stringify([
+      {
+        instanceId: 'Nova',
+        agentConfigId: 'agent-config-nova',
+        guildId: 'guild-001',
+        channelId: 'channel-001',
+        pairedUsers: {
+          'user-001': { role: 'operator', mosaicUserId: 'mosaic-operator-001' },
+        },
+      },
+    ]);
+    // A live session already exists for this conversation/scope, but it was minted under a DIFFERENT
+    // configured agent (Orion). The verified binding reconciles to Nova, so reusing this session would
+    // execute one agent's turn under another agent's verified label — the reuse guard must refuse it.
+    const foreignIdentitySession = {
+      provider: 'configured-provider',
+      modelId: 'configured-model',
+      agentConfigId: 'agent-config-orion',
+      agentName: 'Orion',
+      piSession: {
+        thinkingLevel: 'medium',
+        getAvailableThinkingLevels: (): string[] => ['medium'],
+      },
+    };
+    const prompt = vi.fn().mockResolvedValue(undefined);
+    const createSession = vi.fn().mockResolvedValue(foreignIdentitySession);
+    const agentService = {
+      getSession: vi.fn().mockReturnValue(foreignIdentitySession),
+      createSession,
+      recordMessage: vi.fn(),
+      onEvent: vi.fn().mockReturnValue((): void => undefined),
+      addChannel: vi.fn(),
+      removeChannel: vi.fn(),
+      prompt,
+    };
+    const addMessage = vi.fn().mockResolvedValue({ id: 'discord-persisted-message' });
+    const brain = {
+      agents: { findById: vi.fn((id: string) => Promise.resolve({ id, name: 'Nova' })) },
+      conversations: {
+        findById: vi.fn().mockResolvedValue({ id: 'Nova:discord:channel-001' }),
+        findMessages: vi.fn().mockResolvedValue([]),
+        create: vi.fn().mockResolvedValue(undefined),
+        update: vi.fn().mockResolvedValue(undefined),
+        addMessage,
+      },
+    };
+    const harnessConversations = { append: vi.fn() };
+    const gateway = new ChatGateway(
+      piRpcRouterFronting(agentService, harnessConversations) as never,
+      {} as never,
+      brain as never,
+      {} as never,
+      {} as never,
+      { resolve: vi.fn() } as never,
+    );
+    const client = {
+      id: 'discord-client-identity-swap',
+      data: { discordService: true },
+      emit: vi.fn(),
+    };
+
+    await gateway.handleMessage(
+      client as never,
+      ingressEnvelope('reuse under a different identity', 'discord-identity-swap-001', {
+        conversationId: 'Nova:discord:channel-001',
+      }),
+    );
+
+    // Refused at the embedded reuse guard: no prompt, no persist, no ack — only a typed refusal.
+    expect(prompt).not.toHaveBeenCalled();
+    expect(addMessage).not.toHaveBeenCalled();
+    expect(client.emit).not.toHaveBeenCalledWith('message:ack', expect.anything());
+    expect(client.emit).toHaveBeenCalledWith(
+      'error',
+      expect.objectContaining({ conversationId: 'Nova:discord:channel-001' }),
+    );
+    expect(harnessConversations.append).not.toHaveBeenCalled();
+  });
+
+  it('a verified SEND whose configured agent record resolves under a different id fails reconciliation, consumes no replay claim, and a corrected byte-identical retry dispatches/persists/acks exactly once (Task 5 finding 3 — id axis)', async () => {
+    configureDiscordEnv();
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
+    process.env['DISCORD_ALLOWED_CHANNEL_IDS'] = 'channel-001';
+    process.env['DISCORD_INTERACTION_BINDINGS'] = JSON.stringify([
+      {
+        instanceId: 'Nova',
+        agentConfigId: 'agent-config-nova',
+        guildId: 'guild-001',
+        channelId: 'channel-001',
+        pairedUsers: {
+          'user-001': { role: 'operator', mosaicUserId: 'mosaic-operator-001' },
+        },
+      },
+    ]);
+    const session = {
+      provider: 'configured-provider',
+      modelId: 'configured-model',
+      agentConfigId: 'agent-config-nova',
+      agentName: 'Nova',
+      piSession: {
+        thinkingLevel: 'medium',
+        getAvailableThinkingLevels: (): string[] => ['medium'],
+      },
+    };
+    const createSession = vi.fn().mockResolvedValue(session);
+    const prompt = vi.fn().mockResolvedValue(undefined);
+    const agentService = {
+      getSession: vi.fn().mockReturnValue(undefined),
+      createSession,
+      recordMessage: vi.fn(),
+      onEvent: vi.fn().mockReturnValue((): void => undefined),
+      addChannel: vi.fn(),
+      removeChannel: vi.fn(),
+      prompt,
+    };
+    const addMessage = vi.fn().mockResolvedValue({ id: 'discord-persisted-message' });
+    // The name matches the verified binding, but the record's own id is a DIFFERENT agent config —
+    // an aliased/substituted lookup. Exact-id reconciliation must refuse it on the first delivery,
+    // then admit the corrected record whose id matches the binding.
+    const findAgent = vi
+      .fn()
+      .mockResolvedValueOnce({ id: 'agent-config-elsewhere', name: 'Nova' })
+      .mockResolvedValue({ id: 'agent-config-nova', name: 'Nova' });
+    const brain = {
+      agents: { findById: findAgent },
+      conversations: {
+        findById: vi.fn().mockResolvedValue({ id: 'Nova:discord:channel-001' }),
+        findMessages: vi.fn().mockResolvedValue([]),
+        create: vi.fn().mockResolvedValue(undefined),
+        update: vi.fn().mockResolvedValue(undefined),
+        addMessage,
+      },
+    };
+    const harnessConversations = { append: vi.fn() };
+    const gateway = new ChatGateway(
+      piRpcRouterFronting(agentService, harnessConversations) as never,
+      {} as never,
+      brain as never,
+      {} as never,
+      {} as never,
+      { resolve: vi.fn() } as never,
+    );
+    const client = {
+      id: 'discord-client-reconcile-id',
+      data: { discordService: true },
+      emit: vi.fn(),
+    };
+    const ackCount = (): number =>
+      client.emit.mock.calls.filter((call) => call[0] === 'message:ack').length;
+
+    const envelope = ingressEnvelope(
+      'verified once with aliased agent id',
+      'discord-reconcile-id-001',
+      {
+        conversationId: 'Nova:discord:channel-001',
+      },
+    );
+
+    // (1) The record's id differs from the binding's agentConfigId. Exact-id reconcile refuses the
+    //     turn BEFORE the replay claim, so nothing dispatches and the claim stays available.
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(0);
+    expect(prompt).toHaveBeenCalledTimes(0);
+    expect(addMessage).toHaveBeenCalledTimes(0);
+    expect(ackCount()).toBe(0);
+
+    // (2) The record now reconciles on both id and name; because step (1) took no claim, this
+    //     byte-identical retry claims once and runs the full embedded dispatch exactly once.
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(1);
+    expect(prompt).toHaveBeenCalledTimes(1);
+    expect(addMessage).toHaveBeenCalledTimes(1);
+    expect(ackCount()).toBe(1);
+
+    // (3) A genuine duplicate after the committed turn stays fail-closed.
+    await gateway.handleMessage(client as never, envelope);
+    expect(createSession).toHaveBeenCalledTimes(1);
+    expect(prompt).toHaveBeenCalledTimes(1);
+    expect(addMessage).toHaveBeenCalledTimes(1);
+    expect(ackCount()).toBe(1);
+    expect(harnessConversations.append).not.toHaveBeenCalled();
+  });
+
+  it('a verified SEND refuses a freshly minted same-scope session whose identity differs from the reconciled configured agent, with zero prompt/persist/ack (Task 5 finding 3 — post-create)', async () => {
+    configureDiscordEnv();
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
+    process.env['DISCORD_ALLOWED_CHANNEL_IDS'] = 'channel-001';
+    process.env['DISCORD_INTERACTION_BINDINGS'] = JSON.stringify([
+      {
+        instanceId: 'Nova',
+        agentConfigId: 'agent-config-nova',
+        guildId: 'guild-001',
+        channelId: 'channel-001',
+        pairedUsers: {
+          'user-001': { role: 'operator', mosaicUserId: 'mosaic-operator-001' },
+        },
+      },
+    ]);
+    // No live session exists for this scope, so the runtime MINTS one — but createSession returns a
+    // session carrying a DIFFERENT configured identity (Orion) than the reconciled binding (Nova).
+    // The post-create identity recheck must refuse it rather than dispatch one agent's turn under
+    // another agent's verified label. (The existing reuse test covers the getSession path; this
+    // covers the createSession path scrappy flagged as unvalidated.)
+    const mintedForeignSession = {
+      provider: 'configured-provider',
+      modelId: 'configured-model',
+      agentConfigId: 'agent-config-orion',
+      agentName: 'Orion',
+      piSession: {
+        thinkingLevel: 'medium',
+        getAvailableThinkingLevels: (): string[] => ['medium'],
+      },
+    };
+    const prompt = vi.fn().mockResolvedValue(undefined);
+    const createSession = vi.fn().mockResolvedValue(mintedForeignSession);
+    const agentService = {
+      getSession: vi.fn().mockReturnValue(undefined),
+      createSession,
+      recordMessage: vi.fn(),
+      onEvent: vi.fn().mockReturnValue((): void => undefined),
+      addChannel: vi.fn(),
+      removeChannel: vi.fn(),
+      prompt,
+    };
+    const addMessage = vi.fn().mockResolvedValue({ id: 'discord-persisted-message' });
+    const brain = {
+      agents: { findById: vi.fn((id: string) => Promise.resolve({ id, name: 'Nova' })) },
+      conversations: {
+        findById: vi.fn().mockResolvedValue({ id: 'Nova:discord:channel-001' }),
+        findMessages: vi.fn().mockResolvedValue([]),
+        create: vi.fn().mockResolvedValue(undefined),
+        update: vi.fn().mockResolvedValue(undefined),
+        addMessage,
+      },
+    };
+    const harnessConversations = { append: vi.fn() };
+    const gateway = new ChatGateway(
+      piRpcRouterFronting(agentService, harnessConversations) as never,
+      {} as never,
+      brain as never,
+      {} as never,
+      {} as never,
+      { resolve: vi.fn() } as never,
+    );
+    const client = {
+      id: 'discord-client-postcreate-mismatch',
+      data: { discordService: true },
+      emit: vi.fn(),
+    };
+
+    await gateway.handleMessage(
+      client as never,
+      ingressEnvelope('mint under a different identity', 'discord-postcreate-001', {
+        conversationId: 'Nova:discord:channel-001',
+      }),
+    );
+
+    // The freshly minted session failed the post-create identity recheck: refused with a typed
+    // error, no prompt, no persist, no ack.
+    expect(createSession).toHaveBeenCalledTimes(1);
+    expect(prompt).not.toHaveBeenCalled();
+    expect(addMessage).not.toHaveBeenCalled();
+    expect(client.emit).not.toHaveBeenCalledWith('message:ack', expect.anything());
+    expect(client.emit).toHaveBeenCalledWith(
+      'error',
+      expect.objectContaining({ conversationId: 'Nova:discord:channel-001' }),
+    );
+    expect(harnessConversations.append).not.toHaveBeenCalled();
   });
 
   it('retains validated persisted attachments in resumed conversation history', async () => {
@@ -593,11 +1220,16 @@ describe('Discord ingress security', () => {
 
   it('preserves authenticated attachment metadata through persistence and agent dispatch', async () => {
     configureDiscordEnv();
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
     const prompt = vi.fn().mockResolvedValue(undefined);
-    const addMessage = vi.fn().mockResolvedValue(undefined);
+    const addMessage = vi.fn().mockResolvedValue({ id: 'discord-persisted-message' });
     const session = {
       provider: 'test-provider',
       modelId: 'test-model',
+      // The reused embedded session carries the SAME reconciled identity as the verified binding,
+      // so the finding-3 session-reuse guard admits it rather than refusing an identity swap.
+      agentConfigId: 'agent-config-nova',
+      agentName: 'Nova',
       piSession: {
         thinkingLevel: 'medium',
         getAvailableThinkingLevels: (): string[] => ['medium'],
@@ -611,6 +1243,7 @@ describe('Discord ingress security', () => {
       prompt,
     };
     const brain = {
+      agents: { findById: vi.fn((id: string) => Promise.resolve({ id, name: 'Nova' })) },
       conversations: {
         findById: vi.fn().mockResolvedValue({ id: 'Nova:discord:channel-001' }),
         create: vi.fn().mockResolvedValue(undefined),
@@ -618,8 +1251,9 @@ describe('Discord ingress security', () => {
         addMessage,
       },
     };
+    const harnessConversations = { append: vi.fn() };
     const gateway = new ChatGateway(
-      agentService as never,
+      piRpcRouterFronting(agentService, harnessConversations) as never,
       {} as never,
       brain as never,
       {} as never,
@@ -667,6 +1301,66 @@ describe('Discord ingress security', () => {
       }),
       'discord-service',
     );
+    // The verified Discord prompt dispatch stays on the embedded runtime; the pi-rpc harness that
+    // the router resolved as `active` is never reached.
+    expect(harnessConversations.append).not.toHaveBeenCalled();
+  });
+
+  it('refuses a browser-forged Discord ingress envelope in pi-rpc with a fixed typed refusal and zero dispatch', async () => {
+    // Correction #2 + #4 (behavioural). A browser socket is never `discordService` (that flag is
+    // set only on a valid service-token handshake), so it cannot forge the trusted Discord path by
+    // emitting an envelope-shaped payload. In pi-rpc it must receive a FIXED TYPED refusal
+    // (`runtime_unsupported`, the same typed code the sibling harness-fence uses) and reach neither
+    // the forced Discord service scope, the verified Discord operation, the embedded runtime, nor
+    // the harness. There is no dedicated socket event for verified ingress — the only ingress
+    // surface is the generic `message` handler, and a non-service client is refused there.
+    //
+    // RED today: a non-service client emitting an envelope-shaped payload falls to the browser
+    // branch, fails the chat-message shape check, and is dropped SILENTLY (a warn + return) with no
+    // typed refusal emitted — so the refusal assertion fails. Collection and construction succeed;
+    // the gap is behavioural. GREEN emits the fixed typed refusal before any dispatch.
+    configureDiscordEnv();
+    process.env['CHAT_HARNESS_RUNTIME'] = 'pi-rpc';
+    const agentService = {
+      getSession: vi.fn().mockReturnValue(undefined),
+      createSession: vi.fn(),
+      recordMessage: vi.fn(),
+      onEvent: vi.fn().mockReturnValue((): void => undefined),
+      addChannel: vi.fn(),
+      prompt: vi.fn().mockResolvedValue(undefined),
+    };
+    const harnessConversations = { append: vi.fn() };
+    const routingEngine = { resolve: vi.fn() };
+    const gateway = new ChatGateway(
+      piRpcRouterFronting(agentService, harnessConversations) as never,
+      {} as never,
+      { conversations: { addMessage: vi.fn().mockResolvedValue(undefined) } } as never,
+      {} as never,
+      {} as never,
+      routingEngine as never,
+    );
+    const client = {
+      id: 'browser-forging-discord',
+      data: { discordService: false },
+      emit: vi.fn(),
+    };
+
+    await gateway.handleMessage(
+      client as never,
+      ingressEnvelope('forged from a browser', 'browser-forgery-001', {
+        conversationId: 'Nova:discord:channel-001',
+      }),
+    );
+
+    const refusal = client.emit.mock.calls.find(
+      ([, payload]) => (payload as { code?: string } | undefined)?.code === 'runtime_unsupported',
+    );
+    expect(refusal).toBeDefined();
+    expect(client.emit).not.toHaveBeenCalledWith('message:ack', expect.anything());
+    expect(agentService.createSession).not.toHaveBeenCalled();
+    expect(agentService.prompt).not.toHaveBeenCalled();
+    expect(harnessConversations.append).not.toHaveBeenCalled();
+    expect(routingEngine.resolve).not.toHaveBeenCalled();
   });
 
   it('accepts a thread message through its allowed bound parent channel', () => {

apps/gateway/src/reload/reload.service.spec.ts

@@ -1,5 +1,8 @@
+import { Logger } from '@nestjs/common';
 import { describe, expect, it, vi } from 'vitest';
+import type { SlashCommandPayload, SystemReloadPayload } from '@mosaicstack/types';
 import { ReloadService } from './reload.service.js';
+import { CommandExecutorService } from '../commands/command-executor.service.js';
 
 function createMockCommandRegistry() {
   return {
@@ -104,3 +107,85 @@ describe('ReloadService', () => {
     expect(() => service.registerPlugin('my-plugin', {})).not.toThrow();
   });
 });
+
+describe('ReloadService — /reload command sanitizes plugin errors', () => {
+  it('generic per-plugin errors reach the chat surface while raw markers stay server-side only', async () => {
+    const registry = {
+      getManifest: vi.fn().mockReturnValue({
+        version: 1,
+        commands: [
+          { name: 'reload', aliases: [], scope: 'core', execution: 'socket', available: true },
+        ],
+        skills: [],
+      }),
+    };
+    const reloadService = new ReloadService(registry as never);
+
+    const RELOAD_LOAD_LEAK_MARKER = 'RELOAD_LOAD_LEAK_MARKER /srv/load-secret';
+    const RELOAD_UNLOAD_LEAK_MARKER = 'RELOAD_UNLOAD_LEAK_MARKER /srv/unload-secret';
+
+    reloadService.registerPlugin('unload-fails', {
+      pluginName: 'unload-fails',
+      onLoad: vi.fn().mockResolvedValue(undefined),
+      onUnload: vi.fn().mockRejectedValue(new Error(RELOAD_UNLOAD_LEAK_MARKER)),
+    });
+    reloadService.registerPlugin('load-fails', {
+      pluginName: 'load-fails',
+      onLoad: vi.fn().mockRejectedValue(new Error(RELOAD_LOAD_LEAK_MARKER)),
+      onUnload: vi.fn().mockResolvedValue(undefined),
+    });
+
+    const errorSpy = vi.spyOn(Logger.prototype, 'error').mockImplementation(() => undefined);
+    const broadcastReload = vi.fn();
+    const mockChatGateway = { broadcastReload };
+    const mockAgentService = { getSession: vi.fn(), applyAgentConfig: vi.fn() };
+    const mockSystemOverride = { set: vi.fn(), get: vi.fn(), clear: vi.fn() };
+    const mockSessionGC = { sweepOrphans: vi.fn() };
+    const mockBrain = { agents: { findByName: vi.fn(), findById: vi.fn(), create: vi.fn() } };
+
+    const mockMcpClient = {
+      getServerStatuses: vi.fn(() => []),
+      getToolDefinitions: vi.fn(() => []),
+      reconnectServer: vi.fn().mockResolvedValue(undefined),
+    };
+
+    const executor = new CommandExecutorService(
+      registry as never,
+      mockAgentService as never,
+      mockSystemOverride as never,
+      mockSessionGC as never,
+      null,
+      mockBrain as never,
+      reloadService,
+      mockChatGateway as never,
+      mockMcpClient as never,
+    );
+
+    const payload: SlashCommandPayload = { command: 'reload', conversationId: 'conv-1' };
+    const result = await executor.execute(payload, { userId: 'user-1', tenantId: 'user-1' });
+
+    expect(result.success).toBe(true);
+    expect(result.message).toContain('unload-fails: unload failed (internal error)');
+    expect(result.message).toContain('load-fails: load failed (internal error)');
+    expect(result.message).not.toContain(RELOAD_UNLOAD_LEAK_MARKER);
+    expect(result.message).not.toContain(RELOAD_LOAD_LEAK_MARKER);
+
+    expect(broadcastReload).toHaveBeenCalledOnce();
+    const broadcastPayload = broadcastReload.mock.calls[0]?.[0] as SystemReloadPayload;
+    expect(broadcastPayload.message).toContain('unload-fails: unload failed (internal error)');
+    expect(broadcastPayload.message).toContain('load-fails: load failed (internal error)');
+    expect(broadcastPayload.message).not.toContain(RELOAD_UNLOAD_LEAK_MARKER);
+    expect(broadcastPayload.message).not.toContain(RELOAD_LOAD_LEAK_MARKER);
+
+    const loggedUnloadMarker = errorSpy.mock.calls.some((call) =>
+      call.some((arg) => typeof arg === 'string' && arg.includes(RELOAD_UNLOAD_LEAK_MARKER)),
+    );
+    const loggedLoadMarker = errorSpy.mock.calls.some((call) =>
+      call.some((arg) => typeof arg === 'string' && arg.includes(RELOAD_LOAD_LEAK_MARKER)),
+    );
+    expect(loggedUnloadMarker).toBe(true);
+    expect(loggedLoadMarker).toBe(true);
+
+    errorSpy.mockRestore();
+  });
+});

apps/gateway/src/reload/reload.service.ts

@@ -58,7 +58,8 @@ export class ReloadService implements OnApplicationBootstrap, OnApplicationShutd
           await plugin.onUnload();
           reloaded.push(name);
         } catch (err) {
-          errors.push(`${name}: unload failed — ${err}`);
+          this.logger.error(`Plugin "${name}" failed during onUnload: ${err}`);
+          errors.push(`${name}: unload failed (internal error)`);
         }
       }
     }
@@ -69,7 +70,8 @@ export class ReloadService implements OnApplicationBootstrap, OnApplicationShutd
         try {
           await plugin.onLoad();
         } catch (err) {
-          errors.push(`${name}: load failed — ${err}`);
+          this.logger.error(`Plugin "${name}" failed during onLoad: ${err}`);
+          errors.push(`${name}: load failed (internal error)`);
         }
       }
     }

apps/gateway/src/workspace/workspace.controller.spec.ts

@@ -0,0 +1,104 @@
+import 'reflect-metadata';
+import {
+  type CanActivate,
+  type ExecutionContext,
+  type INestApplication,
+  ValidationPipe,
+} from '@nestjs/common';
+import { FastifyAdapter, type NestFastifyApplication } from '@nestjs/platform-fastify';
+import { Test } from '@nestjs/testing';
+import request from 'supertest';
+import { afterAll, beforeAll, beforeEach, describe, expect, it, vi } from 'vitest';
+import { AuthGuard } from '../auth/auth.guard.js';
+import { ProjectBootstrapService } from './project-bootstrap.service.js';
+import { WorkspaceController } from './workspace.controller.js';
+
+const bootstrapMock = vi.fn(() =>
+  Promise.resolve({
+    projectId: 'project-1',
+    workspacePath: '/opt/mosaic/.workspaces/users/user-1/project-1',
+  }),
+);
+
+const authGuard: CanActivate = {
+  canActivate(context: ExecutionContext): boolean {
+    const requestContext = context.switchToHttp().getRequest<{ user?: { id: string } }>();
+    requestContext.user = { id: 'user-1' };
+    return true;
+  },
+};
+
+describe('POST /api/workspaces repoUrl validation', () => {
+  let app: INestApplication;
+
+  beforeAll(async () => {
+    const moduleRef = await Test.createTestingModule({
+      controllers: [WorkspaceController],
+      providers: [
+        {
+          provide: ProjectBootstrapService,
+          useValue: { bootstrap: bootstrapMock },
+        },
+      ],
+    })
+      .overrideGuard(AuthGuard)
+      .useValue(authGuard)
+      .compile();
+
+    app = moduleRef.createNestApplication<NestFastifyApplication>(new FastifyAdapter());
+    app.useGlobalPipes(
+      new ValidationPipe({
+        whitelist: true,
+        forbidNonWhitelisted: true,
+        transform: true,
+      }),
+    );
+    await app.init();
+    await app.getHttpAdapter().getInstance().ready();
+  });
+
+  beforeEach(() => {
+    bootstrapMock.mockClear();
+  });
+
+  afterAll(async () => {
+    await app.close();
+  });
+
+  it.each([
+    ['a leading-dash value', '--upload-pack=sh -c id'],
+    ['an ext remote helper', 'ext::sh -c id'],
+    ['a file URL', 'file:///tmp/repository'],
+    ['an unparseable value', 'not a url'],
+    ['an SSH shorthand', '[email protected]:acme/repository.git'],
+    ['a scheme without //', 'https:example.com/acme/repository.git'],
+    ['a hostless git URL', 'git:///tmp/repository'],
+  ])('returns 400 for %s', async (_description, repoUrl) => {
+    const response = await request(app.getHttpServer())
+      .post('/api/workspaces')
+      .send({ name: 'Example', repoUrl })
+      .set('Content-Type', 'application/json');
+
+    expect(response.status).toBe(400);
+    expect(bootstrapMock).not.toHaveBeenCalled();
+  });
+
+  it.each([
+    ['a plain HTTPS repository URL', 'https://example.com/acme/repository.git'],
+    ['a git protocol repository URL', 'git://example.com/acme/repository.git'],
+  ])('accepts %s', async (_description, repoUrl) => {
+    const response = await request(app.getHttpServer())
+      .post('/api/workspaces')
+      .send({ name: 'Example', repoUrl })
+      .set('Content-Type', 'application/json');
+
+    expect(response.status).toBe(201);
+    expect(bootstrapMock).toHaveBeenCalledWith({
+      name: 'Example',
+      description: undefined,
+      userId: 'user-1',
+      teamId: undefined,
+      repoUrl,
+    });
+  });
+});

apps/gateway/src/workspace/workspace.controller.ts

@@ -1,7 +1,11 @@
 import { Body, Controller, Post, UseGuards } from '@nestjs/common';
 import { AuthGuard } from '../auth/auth.guard.js';
 import { CurrentUser } from '../auth/current-user.decorator.js';
-import { ProjectBootstrapService } from './project-bootstrap.service.js';
+import {
+  ProjectBootstrapService,
+  type BootstrapProjectResult,
+} from './project-bootstrap.service.js';
+import { CreateWorkspaceDto } from './workspace.dto.js';
 
 @Controller('api/workspaces')
 @UseGuards(AuthGuard)
@@ -11,20 +15,14 @@ export class WorkspaceController {
   @Post()
   async create(
     @CurrentUser() user: { id: string },
-    @Body()
+    @Body() dto: CreateWorkspaceDto,
-    body: {
+  ): Promise<BootstrapProjectResult> {
-      name: string;
-      description?: string;
-      teamId?: string;
-      repoUrl?: string;
-    },
-  ) {
     return this.bootstrap.bootstrap({
-      name: body.name,
+      name: dto.name,
-      description: body.description,
+      description: dto.description,
       userId: user.id,
-      teamId: body.teamId,
+      teamId: dto.teamId,
-      repoUrl: body.repoUrl,
+      repoUrl: dto.repoUrl,
     });
   }
 }

apps/gateway/src/workspace/workspace.dto.ts

@@ -0,0 +1,33 @@
+import { IsOptional, IsString, IsUrl, Matches, MaxLength } from 'class-validator';
+
+export class CreateWorkspaceDto {
+  @IsString()
+  @MaxLength(255)
+  name!: string;
+
+  @IsOptional()
+  @IsString()
+  @MaxLength(10_000)
+  description?: string;
+
+  @IsOptional()
+  @IsString()
+  teamId?: string;
+
+  @IsOptional()
+  @IsString()
+  @Matches(/^(?:https|git):\/\//i, {
+    message: 'repoUrl must be a valid https:// or git:// URL',
+  })
+  @IsUrl(
+    {
+      protocols: ['https', 'git'],
+      require_host: true,
+      require_protocol: true,
+      require_tld: false,
+      require_valid_protocol: true,
+    },
+    { message: 'repoUrl must be a valid https:// or git:// URL' },
+  )
+  repoUrl?: string;
+}

apps/gateway/src/workspace/workspace.service.spec.ts

@@ -1,11 +1,33 @@
-import { describe, it, expect, beforeEach } from 'vitest';
+import { BadRequestException } from '@nestjs/common';
-import { WorkspaceService } from './workspace.service.js';
+import fs from 'node:fs/promises';
+import os from 'node:os';
 import path from 'node:path';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+import { WorkspaceService } from './workspace.service.js';
+
+type ExecFileMock = (
+  command: string,
+  args: readonly string[],
+  options: { cwd: string },
+  callback: (error: Error | null, stdout: string, stderr: string) => void,
+) => void;
+
+const { execFileMock } = vi.hoisted(() => ({
+  execFileMock: vi.fn<ExecFileMock>(),
+}));
+
+vi.mock('node:child_process', () => ({
+  execFile: execFileMock,
+}));
 
 describe('WorkspaceService', () => {
   let service: WorkspaceService;
 
   beforeEach(() => {
+    execFileMock.mockReset();
+    execFileMock.mockImplementation((_command, _args, _options, callback) => {
+      callback(null, '', '');
+    });
     service = new WorkspaceService();
   });
 
@@ -76,4 +98,69 @@ describe('WorkspaceService', () => {
       }
     });
   });
+
+  describe('create', () => {
+    const project = {
+      id: 'project-1',
+      ownerType: 'user',
+      userId: 'user-1',
+      teamId: null,
+    } as const;
+
+    let originalRoot: string | undefined;
+    let temporaryRoot: string;
+
+    beforeEach(async () => {
+      originalRoot = process.env['MOSAIC_ROOT'];
+      temporaryRoot = await fs.mkdtemp(path.join(os.tmpdir(), 'mosaic-workspace-'));
+      process.env['MOSAIC_ROOT'] = temporaryRoot;
+      service = new WorkspaceService();
+    });
+
+    afterEach(async () => {
+      if (originalRoot === undefined) {
+        delete process.env['MOSAIC_ROOT'];
+      } else {
+        process.env['MOSAIC_ROOT'] = originalRoot;
+      }
+      await fs.rm(temporaryRoot, { recursive: true, force: true });
+    });
+
+    it.each([
+      ['a leading-dash URL', '--upload-pack=sh -c id'],
+      ['an ext remote helper', 'ext::sh -c id'],
+      ['a file URL', 'file:///tmp/repository'],
+      ['an unparseable value', 'not a url'],
+      ['an SSH shorthand', '[email protected]:acme/repository.git'],
+      ['a scheme without //', 'https:example.com/acme/repository.git'],
+      ['a hostless git URL', 'git:///tmp/repository'],
+    ])('rejects %s before invoking git', async (_description, repoUrl) => {
+      await expect(service.create(project, repoUrl)).rejects.toBeInstanceOf(BadRequestException);
+      expect(execFileMock).not.toHaveBeenCalled();
+    });
+
+    it.each([
+      ['an HTTPS URL', 'https://example.com/acme/repository.git'],
+      ['a git protocol URL', 'git://example.com/acme/repository.git'],
+    ])('accepts %s and invokes hardened git clone arguments', async (_description, repoUrl) => {
+      const workspacePath = await service.create(project, repoUrl);
+
+      expect(execFileMock).toHaveBeenCalledOnce();
+      expect(execFileMock).toHaveBeenCalledWith(
+        'git',
+        [
+          '-c',
+          'protocol.ext.allow=never',
+          '-c',
+          'protocol.file.allow=never',
+          'clone',
+          '--',
+          repoUrl,
+          '.',
+        ],
+        { cwd: workspacePath },
+        expect.any(Function),
+      );
+    });
+  });
 });

apps/gateway/src/workspace/workspace.service.ts

@@ -1,10 +1,30 @@
-import { Injectable, Logger } from '@nestjs/common';
+import { BadRequestException, Injectable, Logger } from '@nestjs/common';
 import fs from 'node:fs/promises';
 import path from 'node:path';
 import { execFile } from 'node:child_process';
 import { promisify } from 'node:util';
 
 const execFileAsync = promisify(execFile);
+const allowedRepositoryProtocols = new Set(['https:', 'git:']);
+const repositoryUrlPrefixPattern = /^(?:https|git):\/\//i;
+const repositoryUrlError = 'repoUrl must be a valid https:// or git:// URL';
+
+function assertAllowedRepositoryUrl(repoUrl: string): void {
+  if (repoUrl.startsWith('-') || !repositoryUrlPrefixPattern.test(repoUrl)) {
+    throw new BadRequestException(repositoryUrlError);
+  }
+
+  let parsedUrl: URL;
+  try {
+    parsedUrl = new URL(repoUrl);
+  } catch {
+    throw new BadRequestException(repositoryUrlError);
+  }
+
+  if (!allowedRepositoryProtocols.has(parsedUrl.protocol) || parsedUrl.hostname.length === 0) {
+    throw new BadRequestException(repositoryUrlError);
+  }
+}
 
 export interface WorkspaceProject {
   id: string;
@@ -39,14 +59,32 @@ export class WorkspaceService {
    * If repoUrl is provided, clone instead of init.
    */
   async create(project: WorkspaceProject, repoUrl?: string): Promise<string> {
+    if (repoUrl !== undefined) {
+      assertAllowedRepositoryUrl(repoUrl);
+    }
+
     const workspacePath = this.resolvePath(project);
 
     // Create directory
     await fs.mkdir(workspacePath, { recursive: true });
 
-    if (repoUrl) {
+    if (repoUrl !== undefined) {
-      // Clone existing repo
+      // Clone existing repo. Defense in depth keeps dangerous local helpers
-      await execFileAsync('git', ['clone', repoUrl, '.'], { cwd: workspacePath });
+      // disabled and terminates option parsing before positional arguments.
+      await execFileAsync(
+        'git',
+        [
+          '-c',
+          'protocol.ext.allow=never',
+          '-c',
+          'protocol.file.allow=never',
+          'clone',
+          '--',
+          repoUrl,
+          '.',
+        ],
+        { cwd: workspacePath },
+      );
       this.logger.log(`Cloned ${repoUrl} into workspace ${workspacePath}`);
     } else {
       // Init new git repo

apps/web/index.html

@@ -0,0 +1,30 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="UTF-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <title>Mosaic</title>
+    <meta name="description" content="Mosaic Stack Dashboard" />
+    <link rel="preconnect" href="https://fonts.googleapis.com" />
+    <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
+    <link
+      rel="stylesheet"
+      href="https://fonts.googleapis.com/css2?family=Outfit:wght@300;400;500;600;700&family=Fira+Code:wght@400;500&display=swap"
+    />
+    <script>
+      // set data-theme before first paint so the stored theme never flashes
+      (function () {
+        try {
+          var theme = window.localStorage.getItem('mosaic-theme') || 'dark';
+          document.documentElement.setAttribute('data-theme', theme === 'light' ? 'light' : 'dark');
+        } catch (error) {
+          document.documentElement.setAttribute('data-theme', 'dark');
+        }
+      })();
+    </script>
+  </head>
+  <body>
+    <div id="root"></div>
+    <script type="module" src="/src/main.tsx"></script>
+  </body>
+</html>

apps/web/package.json

@@ -4,21 +4,25 @@
   "private": true,
   "scripts": {
     "build": "node ../../scripts/build-web.mjs",
-    "dev": "next dev",
+    "build:vite": "vite build",
+    "dev": "next dev -p 3101",
+    "dev:vite": "vite",
     "lint": "eslint src",
     "typecheck": "tsc --noEmit",
     "test": "vitest run --passWithNoTests",
     "test:e2e": "playwright test",
-    "start": "next start"
+    "start": "next start -p 3101"
   },
   "dependencies": {
     "@mosaicstack/design-tokens": "workspace:^",
+    "@mosaicstack/types": "workspace:^",
     "better-auth": "^1.5.5",
     "clsx": "^2.1.0",
     "next": "^16.0.0",
     "react": "^19.0.0",
     "react-dom": "^19.0.0",
     "react-markdown": "^10.1.0",
+    "react-router-dom": "^7.18.2",
     "socket.io-client": "^4.8.0",
     "tailwind-merge": "^3.5.0"
   },
@@ -28,9 +32,11 @@
     "@types/node": "^22.0.0",
     "@types/react": "^19.0.0",
     "@types/react-dom": "^19.0.0",
+    "@vitejs/plugin-react": "^6.0.5",
     "jsdom": "^29.0.0",
     "tailwindcss": "^4.0.0",
     "typescript": "^5.8.0",
-    "vitest": "^2.0.0"
+    "vite": "^8.2.1",
+    "vitest": "^3.2.7"
   }
 }

apps/web/src/app/auth/provider/[provider]/page.tsx

@@ -3,41 +3,56 @@
 import Link from 'next/link';
 import { useEffect, useState } from 'react';
 import { useParams, useSearchParams } from 'next/navigation';
+import { api } from '@/lib/api';
+import { resolveAuthCallbackURL } from '@/lib/auth-redirect';
 import { signIn } from '@/lib/auth-client';
-import { getSsoProvider } from '@/lib/sso-providers';
+import type { SsoProviderDiscovery } from '@/lib/sso';
 
 export default function AuthProviderRedirectPage(): React.ReactElement {
   const params = useParams<{ provider: string }>();
   const searchParams = useSearchParams();
   const providerId = typeof params.provider === 'string' ? params.provider : '';
-  const provider = getSsoProvider(providerId);
+  const requestedCallbackURL = searchParams.get('callbackURL');
-  const callbackURL = searchParams.get('callbackURL') ?? '/chat';
+  const [providerName, setProviderName] = useState<string | null>(null);
   const [error, setError] = useState<string | null>(null);
 
   useEffect(() => {
-    const currentProvider = provider;
-
-    if (!currentProvider) {
-      setError('Unknown SSO provider.');
-      return;
-    }
-
-    if (!currentProvider.enabled) {
-      setError(`${currentProvider.buttonLabel} is not enabled in this deployment.`);
-      return;
-    }
-
-    const activeProvider = currentProvider;
     let cancelled = false;
 
     async function redirectToProvider(): Promise<void> {
-      const result = await signIn.oauth2({
+      try {
-        providerId: activeProvider.id,
+        const callbackURL = resolveAuthCallbackURL(requestedCallbackURL, window.location.origin);
-        callbackURL,
+        const providers = await api<SsoProviderDiscovery[]>('/api/sso/providers');
-      });
+        if (cancelled) return;
 
-      if (!cancelled && result?.error) {
+        const provider = providers.find((candidate) => candidate.id === providerId);
-        setError(result.error.message ?? `${activeProvider.buttonLabel} sign in failed.`);
+        if (!provider) {
+          setError('Unknown SSO provider.');
+          return;
+        }
+
+        setProviderName(provider.name);
+        if (!provider.configured) {
+          setError(`${provider.name} is not enabled in this deployment.`);
+          return;
+        }
+        if (provider.loginMode !== 'oidc') {
+          setError(`${provider.name} is not available for OIDC sign in.`);
+          return;
+        }
+
+        const result = await signIn.oauth2({
+          providerId: provider.id,
+          callbackURL,
+        });
+
+        if (!cancelled && result?.error) {
+          setError(result.error.message ?? `${provider.name} sign in failed.`);
+        }
+      } catch (caught: unknown) {
+        if (!cancelled) {
+          setError(caught instanceof Error ? caught.message : 'Unable to start single sign-on.');
+        }
       }
     }
 
@@ -46,19 +61,22 @@ export default function AuthProviderRedirectPage(): React.ReactElement {
     return () => {
       cancelled = true;
     };
-  }, [callbackURL, provider]);
+  }, [providerId, requestedCallbackURL]);
 
   return (
     <div className="mx-auto flex min-h-[50vh] max-w-md flex-col justify-center">
       <h1 className="text-2xl font-semibold text-text-primary">Single sign-on</h1>
       <p className="mt-2 text-sm text-text-secondary">
-        {provider
+        {providerName
-          ? `Redirecting you to ${provider.buttonLabel.replace('Continue with ', '')}...`
+          ? `Redirecting you to ${providerName}...`
           : 'Preparing your sign-in request...'}
       </p>
 
       {error ? (
-        <div className="mt-6 rounded-lg border border-error/30 bg-error/10 px-4 py-3 text-sm text-error">
+        <div
+          role="alert"
+          className="mt-6 rounded-lg border border-error/30 bg-error/10 px-4 py-3 text-sm text-error"
+        >
           <p>{error}</p>
           <Link
             href="/login"

apps/web/src/lib/api.spec.ts

@@ -0,0 +1,57 @@
+import { afterEach, describe, expect, it, vi } from 'vitest';
+import { api } from './api';
+
+describe('api', () => {
+  afterEach(() => {
+    vi.unstubAllGlobals();
+  });
+
+  it('fetches the supplied relative path with credentials and a JSON body', async () => {
+    const fetchMock = vi.fn<typeof fetch>();
+    fetchMock.mockResolvedValue(
+      new Response(JSON.stringify({ ok: true }), {
+        status: 200,
+        headers: { 'Content-Type': 'application/json' },
+      }),
+    );
+    vi.stubGlobal('fetch', fetchMock);
+
+    await expect(
+      api<{ ok: boolean }>('/api/projects', {
+        method: 'POST',
+        body: { name: 'Mosaic' },
+      }),
+    ).resolves.toEqual({ ok: true });
+
+    expect(fetchMock).toHaveBeenCalledOnce();
+    expect(fetchMock).toHaveBeenCalledWith(
+      '/api/projects',
+      expect.objectContaining({
+        method: 'POST',
+        credentials: 'include',
+        body: JSON.stringify({ name: 'Mosaic' }),
+        headers: expect.objectContaining({
+          Accept: 'application/json',
+          'Content-Type': 'application/json',
+        }),
+      }),
+    );
+  });
+
+  it('throws the gateway JSON error with its statusCode', async () => {
+    const fetchMock = vi.fn<typeof fetch>();
+    fetchMock.mockResolvedValue(
+      new Response(JSON.stringify({ statusCode: 403, message: 'Forbidden' }), {
+        status: 403,
+        headers: { 'Content-Type': 'application/json' },
+      }),
+    );
+    vi.stubGlobal('fetch', fetchMock);
+
+    await expect(api('/api/admin/users')).rejects.toMatchObject({
+      name: 'Error',
+      message: 'Forbidden',
+      statusCode: 403,
+    });
+  });
+});

apps/web/src/lib/api.ts

@@ -1,5 +1,3 @@
-const GATEWAY_URL = process.env['NEXT_PUBLIC_GATEWAY_URL'] ?? 'http://localhost:14242';
-
 export interface ApiRequestInit extends Omit<RequestInit, 'body'> {
   body?: unknown;
 }
@@ -25,7 +23,7 @@ export async function api<T>(path: string, init?: ApiRequestInit): Promise<T> {
     headers['Content-Type'] = 'application/json';
   }
 
-  const res = await fetch(`${GATEWAY_URL}${path}`, {
+  const res = await fetch(path, {
     credentials: 'include',
     ...rest,
     headers,

apps/web/src/lib/auth-client.spec.ts

@@ -0,0 +1,29 @@
+import { afterEach, describe, expect, it, vi } from 'vitest';
+
+describe('auth client origin contract', () => {
+  afterEach(() => {
+    vi.unstubAllGlobals();
+    vi.resetModules();
+  });
+
+  it('uses the same-origin BetterAuth mount at /api/auth', async () => {
+    const fetchMock = vi.fn<typeof fetch>();
+    fetchMock.mockResolvedValue(
+      new Response(JSON.stringify({ session: null, user: null }), {
+        status: 200,
+        headers: { 'Content-Type': 'application/json' },
+      }),
+    );
+    vi.stubGlobal('fetch', fetchMock);
+
+    const { authClient } = await import('./auth-client');
+    await authClient.getSession();
+
+    expect(fetchMock).toHaveBeenCalledOnce();
+    const firstCall = fetchMock.mock.calls.at(0);
+    expect(firstCall).toBeDefined();
+    const requestURL = new URL(String(firstCall?.[0]), window.location.origin);
+    expect(requestURL.origin).toBe(window.location.origin);
+    expect(requestURL.pathname).toBe('/api/auth/get-session');
+  });
+});

apps/web/src/lib/auth-client.ts

@@ -1,8 +1,9 @@
 import { createAuthClient } from 'better-auth/react';
 import { adminClient, genericOAuthClient } from 'better-auth/client/plugins';
 
+// The gateway and BetterAuth client both use /api/auth. Omitting baseURL keeps
+// every browser request on the current origin in development and production.
 export const authClient = createAuthClient({
-  baseURL: process.env['NEXT_PUBLIC_GATEWAY_URL'] ?? 'http://localhost:14242',
   plugins: [adminClient(), genericOAuthClient()],
 });
 

apps/web/src/lib/auth-redirect.spec.ts

@@ -0,0 +1,30 @@
+import { describe, expect, it } from 'vitest';
+import { resolveAuthCallbackURL } from './auth-redirect';
+
+const CURRENT_ORIGIN = 'https://mosaic.example';
+
+describe('resolveAuthCallbackURL', () => {
+  it('preserves a canonical same-origin path with search and hash', () => {
+    expect(resolveAuthCallbackURL('/projects?view=active#current', CURRENT_ORIGIN)).toBe(
+      '/projects?view=active#current',
+    );
+  });
+
+  it.each([
+    null,
+    'chat',
+    '//evil.example',
+    '/..//evil.com',
+    '/..//evil.com/x',
+    '/./..//evil.com',
+    '/../..//evil.com',
+    '/foo/..//evil.com',
+    '/\\evil.example',
+    '/\n//evil.example',
+    '/\r//evil.example',
+    '/\t//evil.example',
+    'https://evil.example/phish',
+  ])('falls back to chat for an unsafe callback target %#', (candidate) => {
+    expect(resolveAuthCallbackURL(candidate, CURRENT_ORIGIN)).toBe('/chat');
+  });
+});

apps/web/src/lib/auth-redirect.ts

@@ -0,0 +1,23 @@
+const DEFAULT_AUTH_CALLBACK_URL = '/chat';
+
+/**
+ * Return a canonical same-origin path for post-auth navigation.
+ *
+ * Parsing before comparing origins rejects protocol-relative URLs, backslash
+ * variants, and control characters that the WHATWG parser normalizes away.
+ */
+export function resolveAuthCallbackURL(candidate: string | null, currentOrigin: string): string {
+  if (!candidate?.startsWith('/')) return DEFAULT_AUTH_CALLBACK_URL;
+
+  try {
+    const expectedOrigin = new URL(currentOrigin).origin;
+    const resolved = new URL(candidate, expectedOrigin);
+    if (resolved.origin !== expectedOrigin || resolved.pathname.startsWith('//')) {
+      return DEFAULT_AUTH_CALLBACK_URL;
+    }
+
+    return `${resolved.pathname}${resolved.search}${resolved.hash}`;
+  } catch {
+    return DEFAULT_AUTH_CALLBACK_URL;
+  }
+}

apps/web/src/lib/chat-contract.ts

@@ -0,0 +1,71 @@
+// Centralizes the type-only import of the shared `/chat` Socket.IO contract from
+// the public `@mosaicstack/types` package. `import type` is erased at compile
+// time, so this introduces no runtime dependency — it only reuses the exact
+// payload shapes instead of redeclaring them.
+import type { Socket } from 'socket.io-client';
+import type {
+  AbortPayload,
+  AgentEndPayload,
+  AgentStartPayload,
+  AgentTextPayload,
+  AgentThinkingPayload,
+  ChatMessagePayload,
+  ChatSendCapabilityPayload,
+  ChatSendProtocol,
+  ClientToServerEvents,
+  CommandDef,
+  CommandManifest,
+  CommandManifestPayload,
+  ErrorPayload,
+  HarnessSelection,
+  HarnessTurnAckPayload,
+  HarnessTurnSendPayload,
+  MessageAckPayload,
+  RoutingDecisionInfo,
+  ServerToClientEvents,
+  SessionInfoPayload,
+  SessionUsagePayload,
+  SetThinkingPayload,
+  SkillCommandDef,
+  SlashCommandApprovalResultPayload,
+  SlashCommandPayload,
+  SlashCommandResultPayload,
+  SystemReloadPayload,
+  ToolEndPayload,
+  ToolStartPayload,
+} from '@mosaicstack/types';
+
+export type {
+  AbortPayload,
+  AgentEndPayload,
+  AgentStartPayload,
+  AgentTextPayload,
+  AgentThinkingPayload,
+  ChatMessagePayload,
+  ChatSendCapabilityPayload,
+  ChatSendProtocol,
+  ClientToServerEvents,
+  CommandDef,
+  CommandManifest,
+  CommandManifestPayload,
+  ErrorPayload,
+  HarnessSelection,
+  HarnessTurnAckPayload,
+  HarnessTurnSendPayload,
+  MessageAckPayload,
+  RoutingDecisionInfo,
+  ServerToClientEvents,
+  SessionInfoPayload,
+  SessionUsagePayload,
+  SetThinkingPayload,
+  SkillCommandDef,
+  SlashCommandApprovalResultPayload,
+  SlashCommandPayload,
+  SlashCommandResultPayload,
+  SystemReloadPayload,
+  ToolEndPayload,
+  ToolStartPayload,
+};
+
+/** The `/chat` namespace socket, narrowed to the exact typed event contract. */
+export type ChatSocket = Socket<ServerToClientEvents, ClientToServerEvents>;

apps/web/src/lib/socket.spec.ts

@@ -0,0 +1,98 @@
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+
+const { ioMock } = vi.hoisted(() => ({
+  ioMock: vi.fn(),
+}));
+
+vi.mock('socket.io-client', () => ({
+  io: ioMock,
+}));
+
+import { destroySocket, getSocket } from './socket';
+
+interface MockChatSocket {
+  on: ReturnType<typeof vi.fn>;
+  offAny: ReturnType<typeof vi.fn>;
+  disconnect: ReturnType<typeof vi.fn>;
+  /** Test-only helper: fires every handler registered for `event` via
+   * `.on`, mirroring how a real socket.io-client instance invokes its own
+   * listeners (e.g. calling the registered `disconnect` handler(s) on a
+   * real transient disconnect). */
+  trigger(event: string): void;
+}
+
+function createMockSocket(): MockChatSocket {
+  const handlers = new Map<string, Set<() => void>>();
+  const mockSocket: MockChatSocket = {
+    on: vi.fn((event: string, handler: () => void) => {
+      if (!handlers.has(event)) handlers.set(event, new Set());
+      handlers.get(event)?.add(handler);
+      return mockSocket;
+    }),
+    offAny: vi.fn(() => mockSocket),
+    disconnect: vi.fn(() => mockSocket),
+    trigger(event: string): void {
+      for (const handler of handlers.get(event) ?? []) handler();
+    },
+  };
+  return mockSocket;
+}
+
+let currentMock!: MockChatSocket;
+
+describe('chat socket', () => {
+  beforeEach(() => {
+    ioMock.mockReset();
+    // A fresh object per io() call so identity assertions (same singleton vs.
+    // a genuinely new instance) are meaningful.
+    ioMock.mockImplementation(() => {
+      currentMock = createMockSocket();
+      return currentMock;
+    });
+  });
+
+  afterEach(() => {
+    destroySocket();
+  });
+
+  it('creates one same-origin /chat namespace socket', () => {
+    const first = getSocket();
+    const second = getSocket();
+
+    expect(first).toBe(second);
+    expect(ioMock).toHaveBeenCalledOnce();
+    expect(ioMock).toHaveBeenCalledWith('/chat', {
+      withCredentials: true,
+      autoConnect: false,
+      transports: ['websocket', 'polling'],
+    });
+  });
+
+  it('keeps the same singleton instance across a transient disconnect', () => {
+    const first = getSocket();
+
+    // socket.ts must not react to a real socket's `disconnect` event by
+    // nulling the singleton — it registers no such handler at all now.
+    // Actually fire every handler registered via `.on('disconnect', ...)`
+    // (mirroring a real socket.io-client reconnect) instead of merely
+    // calling getSocket() again: this is what makes the test fail if
+    // production reintroduces `socket.on('disconnect', () => { socket =
+    // null; })`, since that handler would run here and null the singleton
+    // before the next getSocket() call.
+    currentMock.trigger('disconnect');
+    const second = getSocket();
+
+    expect(second).toBe(first);
+    expect(ioMock).toHaveBeenCalledOnce();
+  });
+
+  it('only creates a new singleton after an explicit destroySocket()', () => {
+    const first = getSocket();
+
+    destroySocket();
+    const second = getSocket();
+
+    expect(second).not.toBe(first);
+    expect(ioMock).toHaveBeenCalledTimes(2);
+  });
+});

apps/web/src/lib/socket.ts

@@ -1,23 +1,27 @@
-import { io, type Socket } from 'socket.io-client';
+import { io } from 'socket.io-client';
+import type { ChatSocket } from './chat-contract';
 
-const GATEWAY_URL = process.env['NEXT_PUBLIC_GATEWAY_URL'] ?? 'http://localhost:14242';
+let socket: ChatSocket | null = null;
 
-let socket: Socket | null = null;
+export function getSocket(): ChatSocket {
-
-export function getSocket(): Socket {
   if (!socket) {
-    socket = io(`${GATEWAY_URL}/chat`, {
+    // socket.io-client 4.8.3's `io()` factory declaration always returns the
+    // default unparameterized Socket (it accepts no <ListenEvents, EmitEvents>
+    // generics), so this one cast is the unavoidable boundary between that and the
+    // typed `/chat` contract. Every other call site uses the resulting ChatSocket
+    // with no further assertions.
+    socket = io('/chat', {
       withCredentials: true,
       autoConnect: false,
       transports: ['websocket', 'polling'],
-    });
+    }) as unknown as ChatSocket;
 
-    // Reset singleton reference when socket is fully closed so the next
+    // A transient `disconnect` (network blip, server restart) must NOT null
-    // getSocket() call creates a fresh instance instead of returning a
+    // the singleton: socket.io-client auto-reconnects this same instance,
-    // closed/dead socket.
+    // and its listeners stay registered across that reconnect. Nulling here
-    socket.on('disconnect', () => {
+    // previously orphaned those listeners on the next getSocket() call by
-      socket = null;
+    // handing back a brand-new, unconnected instance. Only destroySocket()
-    });
+    // (an explicit, intentional teardown) may reset the singleton.
   }
   return socket;
 }

apps/web/src/lib/sso-providers.test.ts

@@ -1,48 +0,0 @@
-import { afterEach, describe, expect, it, vi } from 'vitest';
-import { getEnabledSsoProviders, getSsoProvider } from './sso-providers';
-
-describe('sso-providers', () => {
-  afterEach(() => {
-    vi.unstubAllEnvs();
-  });
-
-  it('returns the enabled providers in login button order', () => {
-    vi.stubEnv('NEXT_PUBLIC_WORKOS_ENABLED', 'true');
-    vi.stubEnv('NEXT_PUBLIC_KEYCLOAK_ENABLED', 'true');
-
-    expect(getEnabledSsoProviders()).toEqual([
-      {
-        id: 'workos',
-        buttonLabel: 'Continue with WorkOS',
-        description: 'Enterprise SSO via WorkOS',
-        enabled: true,
-        href: '/auth/provider/workos',
-      },
-      {
-        id: 'keycloak',
-        buttonLabel: 'Continue with Keycloak',
-        description: 'Enterprise SSO via Keycloak',
-        enabled: true,
-        href: '/auth/provider/keycloak',
-      },
-    ]);
-  });
-
-  it('marks disabled providers without exposing them in the enabled list', () => {
-    vi.stubEnv('NEXT_PUBLIC_WORKOS_ENABLED', 'true');
-    vi.stubEnv('NEXT_PUBLIC_KEYCLOAK_ENABLED', 'false');
-
-    expect(getEnabledSsoProviders().map((provider) => provider.id)).toEqual(['workos']);
-    expect(getSsoProvider('keycloak')).toEqual({
-      id: 'keycloak',
-      buttonLabel: 'Continue with Keycloak',
-      description: 'Enterprise SSO via Keycloak',
-      enabled: false,
-      href: '/auth/provider/keycloak',
-    });
-  });
-
-  it('returns null for unknown providers', () => {
-    expect(getSsoProvider('authentik')).toBeNull();
-  });
-});

apps/web/src/lib/sso-providers.ts

@@ -1,53 +0,0 @@
-export type SsoProviderId = 'workos' | 'keycloak';
-
-export interface SsoProvider {
-  id: SsoProviderId;
-  buttonLabel: string;
-  description: string;
-  enabled: boolean;
-  href: string;
-}
-
-const PROVIDER_METADATA: Record<SsoProviderId, Omit<SsoProvider, 'enabled' | 'href'>> = {
-  workos: {
-    id: 'workos',
-    buttonLabel: 'Continue with WorkOS',
-    description: 'Enterprise SSO via WorkOS',
-  },
-  keycloak: {
-    id: 'keycloak',
-    buttonLabel: 'Continue with Keycloak',
-    description: 'Enterprise SSO via Keycloak',
-  },
-};
-
-export function getEnabledSsoProviders(): SsoProvider[] {
-  return (Object.keys(PROVIDER_METADATA) as SsoProviderId[])
-    .map((providerId) => getSsoProvider(providerId))
-    .filter((provider): provider is SsoProvider => provider?.enabled === true);
-}
-
-export function getSsoProvider(providerId: string): SsoProvider | null {
-  if (!isSsoProviderId(providerId)) {
-    return null;
-  }
-
-  return {
-    ...PROVIDER_METADATA[providerId],
-    enabled: isSsoProviderEnabled(providerId),
-    href: `/auth/provider/${providerId}`,
-  };
-}
-
-function isSsoProviderId(value: string): value is SsoProviderId {
-  return value === 'workos' || value === 'keycloak';
-}
-
-function isSsoProviderEnabled(providerId: SsoProviderId): boolean {
-  switch (providerId) {
-    case 'workos':
-      return process.env['NEXT_PUBLIC_WORKOS_ENABLED'] === 'true';
-    case 'keycloak':
-      return process.env['NEXT_PUBLIC_KEYCLOAK_ENABLED'] === 'true';
-  }
-}

apps/web/src/lib/types.ts

@@ -1,3 +1,42 @@
+import type {
+  HarnessAuthState,
+  HarnessModelAvailability,
+  HarnessSelection,
+} from '@mosaicstack/types';
+
+// The exact harness/provider/model tuple and its closed enum companions are the
+// shared domain types — re-exported here so web consumers (and the runtime
+// guards) import one shape, never a divergent local redefinition.
+export type { HarnessSelection, HarnessAuthState, HarnessModelAvailability };
+
+/** Harness summary row from `GET /api/harnesses` (the `HarnessSummaryDto`). The
+ * harness id is kept distinct from any provider id — they are never merged. */
+export interface HarnessSummary {
+  id: string;
+  displayName: string;
+  capabilities: string[];
+}
+
+/** One selectable model in a harness catalog. Extends the `{harnessId,
+ * providerId, modelId}` tuple with the display/availability metadata the UI
+ * needs; `inputTypes` is kept as a plain `string[]` on the client boundary
+ * because it arrives from untrusted JSON and is only ever displayed. */
+export interface HarnessCatalogEntry extends HarnessSelection {
+  displayName: string;
+  reasoningCapability: boolean;
+  inputTypes: string[];
+  authState: HarnessAuthState;
+  availability: HarnessModelAvailability;
+}
+
+/** Harness-scoped catalog from `GET /api/harnesses/:harnessId/catalog`. */
+export interface HarnessCatalog {
+  harnessId: string;
+  version: string;
+  fingerprint: string;
+  models: HarnessCatalogEntry[];
+}
+
 /** Conversation returned by the gateway API. */
 export interface Conversation {
   id: string;

apps/web/src/main.tsx

@@ -0,0 +1,19 @@
+import { StrictMode } from 'react';
+import { createRoot } from 'react-dom/client';
+import { RouterProvider } from 'react-router-dom';
+import { ThemeProvider } from '@/providers/theme-provider';
+import { createAppRouter } from '@/routes';
+import '@/app/globals.css';
+
+const container = document.getElementById('root');
+if (!container) {
+  throw new Error('missing #root element');
+}
+
+createRoot(container).render(
+  <StrictMode>
+    <ThemeProvider>
+      <RouterProvider router={createAppRouter()} />
+    </ThemeProvider>
+  </StrictMode>,
+);

apps/web/src/routes.tsx

@@ -0,0 +1,67 @@
+import type { ReactElement } from 'react';
+import { createBrowserRouter, Navigate, Outlet, type RouteObject } from 'react-router-dom';
+import { LoginPage } from '@/spa/pages/login';
+import { RegisterPage } from '@/spa/pages/register';
+import { SsoCallbackPage } from '@/spa/pages/sso-callback';
+import { ChatPage } from '@/spa/pages/chat';
+import { ChatRouteErrorBoundary } from '@/spa/pages/chat-error-boundary';
+import { ProjectDetailPage } from '@/spa/pages/project-detail';
+import { ProjectsPage } from '@/spa/pages/projects';
+import {
+  ProjectDetailRouteErrorBoundary,
+  ProjectsRouteErrorBoundary,
+  TasksRouteErrorBoundary,
+} from '@/spa/pages/resource-route-error-boundaries';
+import { TasksPage } from '@/spa/pages/tasks';
+import { AuthGuard, GuestGuard } from '@/spa/guards';
+import { Placeholder } from '@/spa/placeholder';
+
+function GuestLayout(): ReactElement {
+  return (
+    <div className="flex min-h-screen items-center justify-center bg-surface-bg px-4 py-8">
+      <div className="w-full max-w-md rounded-xl border border-surface-border bg-surface-card p-8 shadow-lg">
+        <Outlet />
+      </div>
+    </div>
+  );
+}
+
+export const routes: RouteObject[] = [
+  {
+    element: <GuestGuard />,
+    children: [
+      {
+        element: <GuestLayout />,
+        children: [
+          { path: '/login', element: <LoginPage /> },
+          { path: '/register', element: <RegisterPage /> },
+          { path: '/auth/provider/:provider', element: <SsoCallbackPage /> },
+        ],
+      },
+    ],
+  },
+  {
+    element: <AuthGuard />,
+    children: [
+      { path: '/', element: <Navigate to="/chat" replace /> },
+      { path: '/chat', element: <ChatPage />, errorElement: <ChatRouteErrorBoundary /> },
+      {
+        path: '/projects',
+        element: <ProjectsPage />,
+        errorElement: <ProjectsRouteErrorBoundary />,
+      },
+      {
+        path: '/projects/:id',
+        element: <ProjectDetailPage />,
+        errorElement: <ProjectDetailRouteErrorBoundary />,
+      },
+      { path: '/tasks', element: <TasksPage />, errorElement: <TasksRouteErrorBoundary /> },
+      { path: '/settings', element: <Placeholder title="Settings" /> },
+      { path: '/admin', element: <Placeholder title="Admin" /> },
+    ],
+  },
+];
+
+export function createAppRouter(): ReturnType<typeof createBrowserRouter> {
+  return createBrowserRouter(routes);
+}

apps/web/src/spa/chat/chat-api.spec.ts

@@ -0,0 +1,195 @@
+import { afterEach, describe, expect, it, vi } from 'vitest';
+import {
+  fetchCatalog,
+  fetchHarnesses,
+  fetchPersistedSelection,
+  persistSelection,
+} from './chat-api';
+
+function json(body: unknown, status = 200): Response {
+  return new Response(JSON.stringify(body), {
+    status,
+    headers: { 'Content-Type': 'application/json' },
+  });
+}
+
+function stubFetch(): ReturnType<typeof vi.fn> {
+  const fetchMock = vi.fn();
+  vi.stubGlobal('fetch', fetchMock);
+  return fetchMock;
+}
+
+/** Every URL the client actually requested, across all calls. */
+function requestedUrls(fetchMock: ReturnType<typeof vi.fn>): string[] {
+  return fetchMock.mock.calls.map((call) => String(call[0]));
+}
+
+describe('chat-api', () => {
+  afterEach(() => {
+    vi.unstubAllGlobals();
+  });
+
+  it('fetchHarnesses GETs /api/harnesses and returns typed summaries (harness id separate from provider)', async () => {
+    const fetchMock = stubFetch();
+    fetchMock.mockResolvedValue(
+      json([
+        { id: 'pi', displayName: 'Pi', capabilities: ['chat', 'tools'] },
+        { id: 'openai', displayName: 'OpenAI', capabilities: ['chat'] },
+      ]),
+    );
+
+    const harnesses = await fetchHarnesses();
+
+    expect(fetchMock).toHaveBeenCalledOnce();
+    expect(String(fetchMock.mock.calls[0]?.[0])).toBe('/api/harnesses');
+    expect(harnesses).toEqual([
+      { id: 'pi', displayName: 'Pi', capabilities: ['chat', 'tools'] },
+      { id: 'openai', displayName: 'OpenAI', capabilities: ['chat'] },
+    ]);
+  });
+
+  it('fetchCatalog GETs the harness-scoped catalog and returns only its model entries', async () => {
+    const fetchMock = stubFetch();
+    fetchMock.mockResolvedValue(
+      json({
+        harnessId: 'pi',
+        version: '2026-08-11',
+        fingerprint: 'abc123',
+        models: [
+          {
+            harnessId: 'pi',
+            providerId: 'openai',
+            modelId: 'gpt-5',
+            displayName: 'GPT-5',
+            reasoningCapability: true,
+            inputTypes: ['text'],
+            authState: 'ready',
+            availability: 'available',
+          },
+        ],
+      }),
+    );
+
+    const result = await fetchCatalog('pi');
+
+    expect(String(fetchMock.mock.calls[0]?.[0])).toBe('/api/harnesses/pi/catalog');
+    expect(result.ok).toBe(true);
+    if (!result.ok) throw new Error('expected ok catalog');
+    expect(result.catalog.harnessId).toBe('pi');
+    expect(result.catalog.models).toHaveLength(1);
+    expect(result.catalog.models[0]).toMatchObject({
+      harnessId: 'pi',
+      providerId: 'openai',
+      modelId: 'gpt-5',
+      availability: 'available',
+    });
+  });
+
+  it('normalizes a catalog 404 into a typed catalog_unavailable result without surfacing the raw body', async () => {
+    const fetchMock = stubFetch();
+    fetchMock.mockResolvedValue(
+      json(
+        {
+          code: 'adapter_unavailable',
+          message: 'raw gateway detail that must not leak verbatim',
+          harnessId: 'attacker-echo',
+          extra: { hostile: 'blob' },
+        },
+        404,
+      ),
+    );
+
+    const result = await fetchCatalog('ghost');
+
+    expect(result.ok).toBe(false);
+    if (result.ok) throw new Error('expected unavailable result');
+    expect(result.code).toBe('catalog_unavailable');
+    // harnessId comes from the request, never the (untrusted) response body.
+    expect(result.harnessId).toBe('ghost');
+    expect(typeof result.message).toBe('string');
+    // The raw response body is never rendered/returned verbatim.
+    expect(JSON.stringify(result)).not.toContain('hostile');
+    expect(JSON.stringify(result)).not.toContain('attacker-echo');
+  });
+
+  it('fetchPersistedSelection returns the stored tuple, or null when unset', async () => {
+    const fetchMock = stubFetch();
+    fetchMock.mockResolvedValueOnce(
+      json({ selection: { harnessId: 'pi', providerId: 'openai', modelId: 'gpt-5' } }),
+    );
+    await expect(fetchPersistedSelection()).resolves.toEqual({
+      harnessId: 'pi',
+      providerId: 'openai',
+      modelId: 'gpt-5',
+    });
+    expect(String(fetchMock.mock.calls[0]?.[0])).toBe('/api/chat/preferences/selection');
+
+    fetchMock.mockResolvedValueOnce(json({ selection: null }));
+    await expect(fetchPersistedSelection()).resolves.toBeNull();
+  });
+
+  it('persistSelection PUTs the structured tuple (not free text) and returns the confirmed selection', async () => {
+    const fetchMock = stubFetch();
+    fetchMock.mockResolvedValue(
+      json({ selection: { harnessId: 'pi', providerId: 'openai', modelId: 'gpt-5' } }),
+    );
+
+    const result = await persistSelection({
+      harnessId: 'pi',
+      providerId: 'openai',
+      modelId: 'gpt-5',
+    });
+
+    expect(result.ok).toBe(true);
+    const call = fetchMock.mock.calls[0];
+    expect(String(call?.[0])).toBe('/api/chat/preferences/selection');
+    const init = call?.[1] as RequestInit;
+    expect(String(init.method).toUpperCase()).toBe('PUT');
+    // The body is exactly the structured tuple — harness/provider/model kept distinct.
+    expect(JSON.parse(String(init.body))).toEqual({
+      harnessId: 'pi',
+      providerId: 'openai',
+      modelId: 'gpt-5',
+    });
+  });
+
+  it('normalizes a selection 422 into a typed error preserving the requested tuple exactly', async () => {
+    const fetchMock = stubFetch();
+    fetchMock.mockResolvedValue(
+      json(
+        {
+          code: 'model_unavailable',
+          message: 'raw detail that must not leak',
+          selection: { harnessId: 'x', providerId: 'y', modelId: 'z' },
+        },
+        422,
+      ),
+    );
+
+    const requested = { harnessId: 'pi', providerId: 'openai', modelId: 'gpt-5' };
+    const result = await persistSelection(requested);
+
+    expect(result.ok).toBe(false);
+    if (result.ok) throw new Error('expected failed persist');
+    expect(['selection_invalid', 'model_unavailable']).toContain(result.code);
+    // The requested tuple is preserved unchanged — not replaced by the body's echo.
+    expect(result.requested).toEqual(requested);
+    expect(JSON.stringify(result)).not.toContain('raw detail');
+  });
+
+  it('never requests any /api/providers* endpoint', async () => {
+    const fetchMock = stubFetch();
+    fetchMock.mockResolvedValue(json([]));
+    await fetchHarnesses();
+    fetchMock.mockResolvedValue(
+      json({ harnessId: 'pi', version: '1', fingerprint: 'f', models: [] }),
+    );
+    await fetchCatalog('pi');
+    fetchMock.mockResolvedValue(json({ selection: null }));
+    await fetchPersistedSelection();
+
+    for (const url of requestedUrls(fetchMock)) {
+      expect(url).not.toContain('/api/providers');
+    }
+  });
+});

apps/web/src/spa/chat/chat-api.ts

@@ -0,0 +1,132 @@
+/**
+ * Typed fetch wrappers for the Task-3 harness HTTP contract the chat selection
+ * UI depends on. Every response body is untrusted and is normalized through the
+ * runtime guards before it reaches state — a 404 (catalog) and a 422 (selection)
+ * are mapped to typed, body-free error results so a raw gateway body is never
+ * rendered, and the caller's requested tuple is preserved verbatim on failure.
+ *
+ * This module talks ONLY to the harness/chat-preferences endpoints. It never
+ * calls `/api/providers*` — provider identity lives inside the harness catalog.
+ */
+import { asHarnessCatalog, asHarnessSelection, asHarnessSummaries } from './runtime-guards';
+import type { HarnessCatalog, HarnessSelection, HarnessSummary } from '@/lib/types';
+
+/** A catalog fetch either yields the typed catalog or a typed unavailability —
+ * never a thrown raw body. */
+export type CatalogResult =
+  | { ok: true; catalog: HarnessCatalog }
+  | { ok: false; code: 'catalog_unavailable'; harnessId: string; message: string };
+
+export type SelectionErrorCode = 'selection_invalid' | 'model_unavailable';
+
+/** A persist either confirms the stored tuple or reports a typed domain failure
+ * that echoes back the exact tuple the caller requested. */
+export type SelectionPersistResult =
+  | { ok: true; selection: HarnessSelection }
+  | { ok: false; code: SelectionErrorCode; message: string; requested: HarnessSelection };
+
+/** A safe, generic message for an unavailable catalog — the raw 404 body is
+ * never surfaced. */
+const CATALOG_UNAVAILABLE_MESSAGE = 'This harness catalog is currently unavailable.';
+
+/** A safe, generic message for a rejected selection. The untrusted 422 body's
+ * own `message` is deliberately NEVER surfaced — only this fixed copy — so a
+ * raw gateway detail can never leak into the UI. Only the closed `code` enum is
+ * read from the body. */
+const SELECTION_REJECTED_MESSAGE = 'This selection was rejected.';
+
+async function readJson(response: Response): Promise<unknown> {
+  return response.json().catch(() => null);
+}
+
+function safeSelectionCode(body: unknown): SelectionErrorCode {
+  if (typeof body === 'object' && body !== null && 'code' in body) {
+    const code = (body as { code: unknown }).code;
+    if (code === 'selection_invalid' || code === 'model_unavailable') return code;
+  }
+  // Default to the more conservative "invalid" classification for anything
+  // unrecognized rather than guessing "model_unavailable".
+  return 'selection_invalid';
+}
+
+/** `GET /api/harnesses` → the list of harness summaries. A non-OK response
+ * normalizes to an empty list (the UI then has no harness to select). */
+export async function fetchHarnesses(): Promise<HarnessSummary[]> {
+  const response = await fetch('/api/harnesses', {
+    credentials: 'include',
+    headers: { Accept: 'application/json' },
+  });
+  if (!response.ok) return [];
+  return asHarnessSummaries(await readJson(response));
+}
+
+/** `GET /api/harnesses/:harnessId/catalog` → the harness-scoped catalog. A 404
+ * (or any non-OK) becomes a typed `catalog_unavailable` result rather than a
+ * fallback catalog or a rendered raw body. */
+export async function fetchCatalog(harnessId: string): Promise<CatalogResult> {
+  const response = await fetch(`/api/harnesses/${encodeURIComponent(harnessId)}/catalog`, {
+    credentials: 'include',
+    headers: { Accept: 'application/json' },
+  });
+  if (!response.ok) {
+    return {
+      ok: false,
+      code: 'catalog_unavailable',
+      // Scoped to the requested harness id, never the untrusted body's echo.
+      harnessId,
+      message: CATALOG_UNAVAILABLE_MESSAGE,
+    };
+  }
+  return { ok: true, catalog: asHarnessCatalog(await readJson(response), harnessId) };
+}
+
+/** `GET /api/chat/preferences/selection` → the persisted tuple, or null when
+ * unset or malformed. */
+export async function fetchPersistedSelection(): Promise<HarnessSelection | null> {
+  const response = await fetch('/api/chat/preferences/selection', {
+    credentials: 'include',
+    headers: { Accept: 'application/json' },
+  });
+  if (!response.ok) return null;
+  const body = await readJson(response);
+  if (typeof body !== 'object' || body === null) return null;
+  return asHarnessSelection((body as { selection?: unknown }).selection);
+}
+
+/** `PUT /api/chat/preferences/selection` with the structured tuple as the body.
+ * On success returns the confirmed selection; on a typed domain failure (422)
+ * or validation error, returns a typed result carrying the EXACT requested
+ * tuple — never the body's echo — and never the raw body text. */
+export async function persistSelection(
+  selection: HarnessSelection,
+): Promise<SelectionPersistResult> {
+  const requested: HarnessSelection = {
+    harnessId: selection.harnessId,
+    providerId: selection.providerId,
+    modelId: selection.modelId,
+  };
+  const response = await fetch('/api/chat/preferences/selection', {
+    method: 'PUT',
+    credentials: 'include',
+    headers: { Accept: 'application/json', 'Content-Type': 'application/json' },
+    body: JSON.stringify(requested),
+  });
+  if (!response.ok) {
+    const body = await readJson(response);
+    return {
+      ok: false,
+      code: safeSelectionCode(body),
+      // Fixed copy only — the untrusted body's message is never surfaced.
+      message: SELECTION_REJECTED_MESSAGE,
+      requested,
+    };
+  }
+  const body = await readJson(response);
+  const confirmed =
+    typeof body === 'object' && body !== null
+      ? asHarnessSelection((body as { selection?: unknown }).selection)
+      : null;
+  // A malformed 2xx body is treated as a confirmation of exactly what we sent —
+  // the server accepted the tuple, so the requested tuple is the source of truth.
+  return { ok: true, selection: confirmed ?? requested };
+}

apps/web/src/spa/chat/commands-panel.spec.tsx

@@ -0,0 +1,280 @@
+import { act } from 'react';
+import { createRoot, type Root } from 'react-dom/client';
+import { afterAll, afterEach, beforeAll, describe, expect, it, vi } from 'vitest';
+import { CommandsPanel } from './commands-panel';
+
+beforeAll(() => {
+  Object.defineProperty(globalThis, 'IS_REACT_ACT_ENVIRONMENT', {
+    configurable: true,
+    value: true,
+  });
+});
+
+afterAll(() => {
+  Reflect.deleteProperty(globalThis, 'IS_REACT_ACT_ENVIRONMENT');
+});
+
+let root: Root | null;
+let container: HTMLElement | null;
+
+async function render(node: Parameters<Root['render']>[0]): Promise<void> {
+  container = document.createElement('div');
+  document.body.append(container);
+  root = createRoot(container);
+  await act(async () => {
+    root?.render(node);
+  });
+}
+
+afterEach(async () => {
+  await act(async () => {
+    root?.unmount();
+  });
+  document.body.replaceChildren();
+  root = null;
+  container = null;
+});
+
+describe('CommandsPanel', () => {
+  it('shows the frozen local pendingApproval args in the confirmation area, regardless of misleading server message text', async () => {
+    await render(
+      <CommandsPanel
+        manifest={null}
+        results={[]}
+        approval={{
+          conversationId: 'c1',
+          command: 'deploy', // matches pendingApproval — this is a legitimately approved request
+          success: true,
+          approvalId: 'ap1',
+          expiresAt: '2026-01-01T00:00:00.000Z',
+          // Free-text server message claims a different, less alarming target
+          // than what will actually be sent — the UI must not rely on this.
+          message: 'This will only affect the staging environment.',
+        }}
+        pendingApproval={{ command: 'deploy', args: 'prod' }}
+        hasConversation
+        onExecute={vi.fn()}
+        onApprove={vi.fn()}
+        onRunApproved={vi.fn()}
+      />,
+    );
+
+    // The exact frozen combined action is visible...
+    expect(container?.textContent).toContain('/deploy');
+    expect(container?.textContent).toContain('prod');
+    // ...and the misleading server free-text is never shown next to it.
+    expect(container?.textContent).not.toContain('staging environment');
+  });
+
+  it('does not throw when a manifest commands entry is null', async () => {
+    const manifest = {
+      commands: [
+        null,
+        {
+          name: 'model',
+          aliases: [],
+          description: 'Change the active model',
+          scope: 'core',
+          execution: 'socket',
+          available: true,
+        },
+      ],
+      skills: [null],
+      version: 1,
+    } as unknown as Parameters<typeof CommandsPanel>[0]['manifest'];
+
+    await expect(
+      render(
+        <CommandsPanel
+          manifest={manifest}
+          results={[]}
+          approval={null}
+          pendingApproval={null}
+          hasConversation={false}
+          onExecute={vi.fn()}
+          onApprove={vi.fn()}
+          onRunApproved={vi.fn()}
+        />,
+      ),
+    ).resolves.not.toThrow();
+
+    expect(container?.textContent).toContain('model');
+  });
+
+  it('shows an explicit no-args fallback when the frozen pendingApproval has no args', async () => {
+    await render(
+      <CommandsPanel
+        manifest={null}
+        results={[]}
+        approval={{
+          conversationId: 'c1',
+          command: 'deploy',
+          success: true,
+          approvalId: 'ap1',
+          expiresAt: '2026-01-01T00:00:00.000Z',
+        }}
+        pendingApproval={{ command: 'deploy' }}
+        hasConversation
+        onExecute={vi.fn()}
+        onApprove={vi.fn()}
+        onRunApproved={vi.fn()}
+      />,
+    );
+
+    expect(container?.textContent?.toLowerCase()).toContain('no args');
+  });
+
+  it('renders skills from a skills-only manifest', async () => {
+    await render(
+      <CommandsPanel
+        manifest={{
+          commands: [],
+          skills: [{ name: 'brave-search', description: 'Search the web', available: true }],
+          version: 1,
+        }}
+        results={[]}
+        approval={null}
+        pendingApproval={null}
+        hasConversation={false}
+        onExecute={vi.fn()}
+        onApprove={vi.fn()}
+        onRunApproved={vi.fn()}
+      />,
+    );
+
+    expect(container?.textContent).toContain('brave-search');
+    expect(container?.textContent).toContain('Search the web');
+  });
+
+  it('does not show the Run affordance when approval.success/approvalId are objects, even though command matches pendingApproval', async () => {
+    const approval = {
+      conversationId: 'c1',
+      command: 'deploy',
+      success: { truthy: 'object' },
+      approvalId: { also: 'object' },
+    } as unknown as Parameters<typeof CommandsPanel>[0]['approval'];
+
+    await render(
+      <CommandsPanel
+        manifest={null}
+        results={[]}
+        approval={approval}
+        pendingApproval={{ command: 'deploy', args: 'prod' }}
+        hasConversation
+        onExecute={vi.fn()}
+        onApprove={vi.fn()}
+        onRunApproved={vi.fn()}
+      />,
+    );
+
+    expect(
+      [...(container?.querySelectorAll('button') ?? [])].some((button) =>
+        button.textContent?.includes('Run approved command'),
+      ),
+    ).toBe(false);
+  });
+
+  it('shows the guarded server-provided denial reason for a denied approval', async () => {
+    await render(
+      <CommandsPanel
+        manifest={null}
+        results={[]}
+        approval={{
+          conversationId: 'c1',
+          command: 'deploy',
+          success: false,
+          message: 'Not authorized',
+        }}
+        pendingApproval={{ command: 'deploy', args: 'prod' }}
+        hasConversation
+        onExecute={vi.fn()}
+        onApprove={vi.fn()}
+        onRunApproved={vi.fn()}
+      />,
+    );
+
+    expect(container?.textContent).toContain('Not authorized');
+  });
+
+  it('falls back to a stable "Denied." copy when a denial has no usable message', async () => {
+    await render(
+      <CommandsPanel
+        manifest={null}
+        results={[]}
+        approval={{ conversationId: 'c1', command: 'deploy', success: false }}
+        pendingApproval={{ command: 'deploy', args: 'prod' }}
+        hasConversation
+        onExecute={vi.fn()}
+        onApprove={vi.fn()}
+        onRunApproved={vi.fn()}
+      />,
+    );
+
+    expect(container?.textContent).toContain('Denied.');
+  });
+
+  it('shows the guarded contract-provided reason for a failed command result, falling back to a stable copy only when absent', async () => {
+    await render(
+      <CommandsPanel
+        manifest={null}
+        results={[
+          { conversationId: 'c1', command: 'model', success: false, message: 'Unknown model' },
+          { conversationId: 'c1', command: 'deploy', success: false },
+        ]}
+        approval={null}
+        pendingApproval={null}
+        hasConversation={false}
+        onExecute={vi.fn()}
+        onApprove={vi.fn()}
+        onRunApproved={vi.fn()}
+      />,
+    );
+
+    expect(container?.textContent).toContain('Unknown model');
+    expect(container?.textContent).toContain('Command failed.');
+  });
+
+  it('bounds an oversized command result message at the render site as defense-in-depth', async () => {
+    const hostileMessage = 'y'.repeat(50_000);
+    await render(
+      <CommandsPanel
+        manifest={null}
+        results={[
+          { conversationId: 'c1', command: 'model', success: false, message: hostileMessage },
+        ]}
+        approval={null}
+        pendingApproval={null}
+        hasConversation={false}
+        onExecute={vi.fn()}
+        onApprove={vi.fn()}
+        onRunApproved={vi.fn()}
+      />,
+    );
+
+    const text = container?.textContent ?? '';
+    expect(text.length).toBeLessThan(hostileMessage.length);
+  });
+
+  it('does not throw when the manifest fields are malformed (non-array commands/skills)', async () => {
+    const manifest = {
+      commands: 'not-an-array',
+      skills: null,
+      version: 1,
+    } as unknown as Parameters<typeof CommandsPanel>[0]['manifest'];
+
+    await expect(
+      render(
+        <CommandsPanel
+          manifest={manifest}
+          results={[]}
+          approval={null}
+          pendingApproval={null}
+          hasConversation={false}
+          onExecute={vi.fn()}
+          onApprove={vi.fn()}
+          onRunApproved={vi.fn()}
+        />,
+      ),
+    ).resolves.not.toThrow();
+  });
+});

apps/web/src/spa/chat/commands-panel.tsx

@@ -0,0 +1,164 @@
+import { useState, type ReactElement } from 'react';
+import type { PendingApproval } from './use-chat-connection';
+import { MAX_COMMAND_MESSAGE_CHARS } from './limits';
+import { asNonEmptyString, asString } from './runtime-guards';
+import type {
+  CommandManifest,
+  SlashCommandApprovalResultPayload,
+  SlashCommandResultPayload,
+} from '@/lib/chat-contract';
+
+/** Stable fallback copy shown for a failed command only when the server's
+ * own guarded, non-empty `message` (e.g. "Unknown model") is absent or
+ * malformed — the structured contract reason itself is otherwise shown
+ * directly, never a raw thrown exception, stack trace, or object value. */
+const COMMAND_FAILURE_COPY = 'Command failed.';
+
+/** Render-site defense-in-depth: `use-chat-connection.ts` already bounds a
+ * stored command:result message at ingestion, but this component must never
+ * assume every caller went through that path — bounding again here means a
+ * hostile/oversized message can never force an unbounded render. */
+function boundMessage(value: string): string {
+  return value.length > MAX_COMMAND_MESSAGE_CHARS
+    ? value.slice(0, MAX_COMMAND_MESSAGE_CHARS)
+    : value;
+}
+
+interface CommandsPanelProps {
+  manifest: CommandManifest | null;
+  results: SlashCommandResultPayload[];
+  approval: SlashCommandApprovalResultPayload | null;
+  pendingApproval: PendingApproval | null;
+  hasConversation: boolean;
+  onExecute: (input: { command: string; args?: string }) => void;
+  onApprove: (input: { command: string; args?: string }) => void;
+  onRunApproved: () => void;
+}
+
+export function CommandsPanel({
+  manifest,
+  results,
+  approval,
+  pendingApproval,
+  hasConversation,
+  onExecute,
+  onApprove,
+  onRunApproved,
+}: CommandsPanelProps): ReactElement {
+  const [command, setCommand] = useState('');
+  const [args, setArgs] = useState('');
+
+  // Defense-in-depth: the reducer already normalizes success/approvalId
+  // before storing `approval`, but a matching command string alone must
+  // never be trusted here either — require the literal boolean `true` and a
+  // non-empty string approvalId, not merely truthy values.
+  const canRunApproved =
+    approval?.success === true &&
+    typeof approval.approvalId === 'string' &&
+    approval.approvalId.length > 0 &&
+    !!pendingApproval &&
+    pendingApproval.command === approval.command;
+
+  // A manifest arrives from the server as untyped JSON at runtime — guard
+  // both collections before mapping so a malformed manifest cannot throw.
+  const commands = Array.isArray(manifest?.commands) ? manifest.commands : [];
+  const skills = Array.isArray(manifest?.skills) ? manifest.skills : [];
+
+  return (
+    <section aria-label="Commands" className="flex flex-col gap-2 border-b px-4 py-3 text-xs">
+      {commands.length > 0 ? (
+        <ul aria-label="Available commands" className="flex flex-col gap-1">
+          {commands.map((cmd, index) => (
+            <li key={asString(cmd?.name) || `cmd-${index}`}>
+              <strong>/{asString(cmd?.name)}</strong> — {asString(cmd?.description)}
+            </li>
+          ))}
+        </ul>
+      ) : null}
+
+      {skills.length > 0 ? (
+        <ul aria-label="Available skills" className="flex flex-col gap-1">
+          {skills.map((skill, index) => (
+            <li key={asString(skill?.name) || `skill-${index}`}>
+              <strong>/skill:{asString(skill?.name)}</strong> — {asString(skill?.description)}
+            </li>
+          ))}
+        </ul>
+      ) : null}
+
+      <div className="flex flex-wrap items-center gap-2">
+        <input
+          aria-label="Command name"
+          value={command}
+          onChange={(event) => setCommand(event.target.value)}
+          placeholder="command"
+        />
+        <input
+          aria-label="Command arguments"
+          value={args}
+          onChange={(event) => setArgs(event.target.value)}
+          placeholder="args (optional)"
+        />
+        <button
+          type="button"
+          disabled={!hasConversation || !command.trim()}
+          onClick={() => onExecute({ command: command.trim(), args: args.trim() || undefined })}
+        >
+          Run command
+        </button>
+        <button
+          type="button"
+          disabled={!hasConversation || !command.trim()}
+          onClick={() => onApprove({ command: command.trim(), args: args.trim() || undefined })}
+        >
+          Request approval
+        </button>
+      </div>
+
+      {approval ? (
+        <div role={approval.success ? 'status' : 'alert'} className="flex items-center gap-2">
+          {/* A successful approval shows stable client copy only — never
+              the server-controlled approval.message or echoed
+              approval.command as the primary confirmation. The frozen local
+              pendingApproval below (not this line) is the sole authoritative
+              statement of what will run. A denial, by contrast, is not an
+              execution authority and safely surfaces the guarded structured
+              reason the server gave (e.g. "Not authorized"), falling back to
+              a stable copy only when absent/malformed. */}
+          <span>
+            {approval.success ? 'Approved.' : asNonEmptyString(approval.message, 'Denied.')}
+          </span>
+          {canRunApproved && pendingApproval ? (
+            <>
+              {/* Authoritative frozen local command+args — what the click below
+                  will actually emit. The server's `approval` above is display-only
+                  and must never be trusted to represent the executed payload. */}
+              <span>
+                Will run: /{pendingApproval.command}{' '}
+                {pendingApproval.args ? pendingApproval.args : '(no args)'}
+              </span>
+              <button type="button" onClick={onRunApproved}>
+                Run approved command
+              </button>
+            </>
+          ) : null}
+        </div>
+      ) : null}
+
+      {results.length > 0 ? (
+        <ul aria-label="Command results" className="flex flex-col gap-1">
+          {results.map((result, index) => (
+            <li key={`${result.command}-${index}`} role={result.success ? 'status' : 'alert'}>
+              /{asString(result.command)}: {result.success ? 'success' : 'failed'}
+              {result.success
+                ? typeof result.message === 'string' && result.message
+                  ? ` — ${boundMessage(result.message)}`
+                  : ''
+                : ` — ${boundMessage(asNonEmptyString(result.message, COMMAND_FAILURE_COPY))}`}
+            </li>
+          ))}
+        </ul>
+      ) : null}
+    </section>
+  );
+}

apps/web/src/spa/chat/composer.tsx

@@ -0,0 +1,181 @@
+import { useState, type KeyboardEvent, type ReactElement } from 'react';
+import type { HarnessSelection } from '@/lib/types';
+import type { HarnessSelectionValue } from './use-harness-selection';
+
+interface ComposerProps {
+  onSend: (input: { content: string; selection: HarnessSelection }) => boolean;
+  onStop: () => void;
+  streaming: boolean;
+  /** True from local send time through server turn startup/ack and
+   * throughout streaming — a superset of `streaming` that also covers the
+   * pre-ack window where a second send could otherwise slip through. */
+  sending: boolean;
+  hasConversation: boolean;
+  /** Structured harness/provider/model selection state. The composer never
+   * accepts free-text provider/model — every sendable tuple is a validated,
+   * persisted catalog entry, and the send projection is derived from it. */
+  harness: HarnessSelectionValue;
+}
+
+/** The distinct provider ids present in the current catalog, in first-seen
+ * order — the provider select is catalog-derived, never a hardcoded list. */
+function providerOptions(harness: HarnessSelectionValue): string[] {
+  const seen = new Set<string>();
+  const out: string[] = [];
+  for (const model of harness.catalog?.models ?? []) {
+    if (seen.has(model.providerId)) continue;
+    seen.add(model.providerId);
+    out.push(model.providerId);
+  }
+  return out;
+}
+
+export function Composer({
+  onSend,
+  onStop,
+  streaming,
+  sending,
+  hasConversation,
+  harness,
+}: ComposerProps): ReactElement {
+  const [content, setContent] = useState('');
+  const busy = streaming || sending;
+
+  function submit(): void {
+    if (busy) return;
+    // Send is gated on a validated, persisted catalog tuple — a draft or unset
+    // selection can never emit, so provider/model never travel as free text.
+    if (!harness.canSend || harness.persistedSelection === null) return;
+    const trimmed = content.trim();
+    if (!trimmed) return;
+    // Pass the validated, persisted selection tuple only. The hook derives the
+    // wire projection (legacy `message` provider/model, or `turn:send`) from the
+    // negotiated `chat:send-capability` protocol — never from flat caller input.
+    const selection = harness.persistedSelection;
+    const ok = onSend({ content: trimmed, selection });
+    // Clear the input only when the send was accepted — a refused turn (e.g. a
+    // failed idempotency mint) must retain the user's text so it is not lost.
+    if (ok) setContent('');
+  }
+
+  function handleKeyDown(event: KeyboardEvent<HTMLTextAreaElement>): void {
+    if (event.key === 'Enter' && !event.shiftKey) {
+      event.preventDefault();
+      submit();
+    }
+  }
+
+  // Scope the model options to the intentionally selected provider. With no
+  // provider chosen (`providerId === ''`) nothing matches, so the model select
+  // offers only the placeholder — never a cross-provider row.
+  const models = (harness.catalog?.models ?? []).filter(
+    (model) => model.providerId === harness.providerId,
+  );
+  // A collision-safe composite option identity covering the full provider+model
+  // tuple. The controlled select mirrors the same identity so the exact catalog
+  // row highlights (a bare modelId would collide across providers).
+  const modelOptionValue = (model: { providerId: string; modelId: string }): string =>
+    `${model.providerId}:${model.modelId}`;
+  const selectedModelValue = harness.modelId ? `${harness.providerId}:${harness.modelId}` : '';
+
+  return (
+    <form
+      onSubmit={(event) => {
+        event.preventDefault();
+        submit();
+      }}
+      className="flex flex-col gap-2 border-t p-4"
+    >
+      <div className="flex flex-wrap gap-2">
+        <select
+          aria-label="Harness"
+          value={harness.harnessId}
+          onChange={(event) => harness.selectHarness(event.target.value)}
+          className="rounded border px-2 py-1 text-xs"
+        >
+          <option value="">Select a harness…</option>
+          {harness.harnesses.map((item) => (
+            <option key={item.id} value={item.id}>
+              {item.displayName}
+            </option>
+          ))}
+        </select>
+        <select
+          aria-label="Provider"
+          value={harness.providerId}
+          onChange={(event) => harness.selectProvider(event.target.value)}
+          disabled={harness.catalogUnavailable || providerOptions(harness).length === 0}
+          className="rounded border px-2 py-1 text-xs"
+        >
+          <option value="">Select a provider…</option>
+          {providerOptions(harness).map((providerId) => (
+            <option key={providerId} value={providerId}>
+              {providerId}
+            </option>
+          ))}
+        </select>
+        <select
+          aria-label="Model"
+          value={selectedModelValue}
+          onChange={(event) => {
+            // Resolve the composite option identity back to the exact catalog
+            // row and persist that row's own provider+model — never a bare id.
+            const selected = models.find((model) => modelOptionValue(model) === event.target.value);
+            if (selected) harness.selectModel(selected.providerId, selected.modelId);
+          }}
+          disabled={harness.catalogUnavailable || models.length === 0}
+          className="rounded border px-2 py-1 text-xs"
+        >
+          <option value="">Select a model…</option>
+          {models.map((model) => (
+            <option key={modelOptionValue(model)} value={modelOptionValue(model)}>
+              {model.displayName}
+            </option>
+          ))}
+        </select>
+      </div>
+      {harness.catalogUnavailable ? (
+        <p role="status" className="text-xs opacity-70">
+          This harness catalog is currently unavailable.
+        </p>
+      ) : null}
+      {harness.isStale ? (
+        <p role="status" className="text-xs opacity-70">
+          The saved model is no longer available — pick another to continue.
+        </p>
+      ) : null}
+      {harness.persistError ? (
+        <p role="alert" className="text-xs">
+          {harness.persistError.message}
+        </p>
+      ) : null}
+      <div className="flex items-end gap-2">
+        <textarea
+          aria-label="Message"
+          value={content}
+          onChange={(event) => setContent(event.target.value)}
+          onKeyDown={handleKeyDown}
+          rows={2}
+          placeholder="Message… (Enter to send, Shift+Enter for a new line)"
+          className="flex-1 resize-none rounded border px-3 py-2 text-sm"
+        />
+        <button
+          type="submit"
+          disabled={!content.trim() || busy || !harness.canSend}
+          className="rounded px-3 py-2 text-sm font-medium"
+        >
+          Send
+        </button>
+        <button
+          type="button"
+          aria-label="Stop"
+          disabled={!hasConversation || !streaming}
+          onClick={onStop}
+          className="rounded px-3 py-2 text-sm font-medium"
+        >
+          Stop
+        </button>
+      </div>
+    </form>
+  );
+}

apps/web/src/spa/chat/limits.ts

@@ -0,0 +1,22 @@
+/**
+ * Bounds on server-fed chat state. A hostile or malfunctioning gateway can
+ * flood any of these collections; caps keep memory/render cost flat instead
+ * of growing unboundedly for the lifetime of the connection.
+ */
+
+/** Max characters retained for the in-flight streamed text/thinking buffers. */
+export const MAX_STREAM_CHARS = 20_000;
+/** Max transcript turns retained (oldest dropped first). */
+export const MAX_MESSAGES = 500;
+/** Max tool-call entries (including anomaly entries) retained per turn history. */
+export const MAX_TOOLS = 200;
+/** Max slash-command results retained. */
+export const MAX_COMMAND_RESULTS = 200;
+/** Max commands/skills accepted from a single manifest push. */
+export const MAX_MANIFEST_ITEMS = 500;
+/** Max executed approval IDs remembered for single-flight dedup. */
+export const MAX_EXECUTED_APPROVAL_IDS = 200;
+/** Max characters retained for a single command:result message — a hostile
+ * or malfunctioning gateway must not be able to push an unbounded curated
+ * success/failure reason into state (or, defensively, onto the page). */
+export const MAX_COMMAND_MESSAGE_CHARS = 1_000;

apps/web/src/spa/chat/message-transcript.tsx

@@ -0,0 +1,39 @@
+import type { ReactElement } from 'react';
+import type { ChatTranscriptMessage } from './use-chat-connection';
+
+interface MessageTranscriptProps {
+  messages: ChatTranscriptMessage[];
+  streaming: boolean;
+  text: string;
+}
+
+export function MessageTranscript({
+  messages,
+  streaming,
+  text,
+}: MessageTranscriptProps): ReactElement {
+  return (
+    <div
+      role="log"
+      aria-live="polite"
+      aria-label="Conversation"
+      className="flex flex-1 flex-col gap-3 overflow-y-auto p-4"
+    >
+      {messages.map((message) => (
+        <div key={message.id} data-role={message.role} className="whitespace-pre-wrap text-sm">
+          <span className="font-medium">{message.role === 'user' ? 'You' : 'Assistant'}: </span>
+          <span>{message.text}</span>
+          {message.thinking ? (
+            <div className="pt-1 text-xs italic opacity-70">{message.thinking}</div>
+          ) : null}
+        </div>
+      ))}
+      {streaming ? (
+        <div data-role="assistant-streaming" className="whitespace-pre-wrap text-sm">
+          <span className="font-medium">Assistant: </span>
+          <span>{text || 'Thinking…'}</span>
+        </div>
+      ) : null}
+    </div>
+  );
+}

apps/web/src/spa/chat/runtime-guards.ts

@@ -0,0 +1,150 @@
+/**
+ * Socket.IO payloads are only statically typed at the call site — a
+ * misbehaving or compromised gateway can send anything at runtime. These
+ * guards protect the dereference sites that would otherwise throw (`.map` on
+ * a non-array, `.toFixed` on a non-number) or render an object as a React
+ * child.
+ */
+import type {
+  HarnessAuthState,
+  HarnessCatalog,
+  HarnessCatalogEntry,
+  HarnessModelAvailability,
+  HarnessSelection,
+  HarnessSummary,
+} from '@/lib/types';
+
+export function asString(value: unknown, fallback = ''): string {
+  return typeof value === 'string' ? value : fallback;
+}
+
+/** Like `asString`, but an empty string also falls back — used for guarded
+ * contract-provided reason strings (e.g. a denial or failure message) where
+ * an empty string is not a meaningful value to display in place of the
+ * stable fallback copy. */
+export function asNonEmptyString(value: unknown, fallback: string): string {
+  return typeof value === 'string' && value.length > 0 ? value : fallback;
+}
+
+export function asFiniteNumber(value: unknown, fallback = 0): number {
+  return typeof value === 'number' && Number.isFinite(value) ? value : fallback;
+}
+
+/** Like `asFiniteNumber`, but returns `null` on failure instead of a numeric
+ * fallback — callers that must not fabricate a plausible-looking value (e.g.
+ * `0 tokens` / `$0.0000` for genuinely unknown usage) use this to render an
+ * honest "unavailable" label instead. */
+export function asFiniteNumberOrNull(value: unknown): number | null {
+  return typeof value === 'number' && Number.isFinite(value) ? value : null;
+}
+
+export function asStringArray(value: unknown): string[] {
+  return Array.isArray(value) && value.every((item) => typeof item === 'string') ? value : [];
+}
+
+export function isRecord(value: unknown): value is Record<string, unknown> {
+  return typeof value === 'object' && value !== null;
+}
+
+/**
+ * The HTTP harness/catalog/selection JSON bodies are as untrusted as the socket
+ * payloads above — a misbehaving or compromised gateway can send anything. The
+ * guards below normalize those bodies into the typed client shapes without ever
+ * rendering a raw body, so a 404/422/malformed response can never inject an
+ * object into React or a non-tuple into the selection state.
+ */
+
+/** Normalizes an untrusted `authState` to the closed set, defaulting to the
+ * safest value (`unavailable`) for anything unrecognized. */
+export function asHarnessAuthState(value: unknown): HarnessAuthState {
+  return value === 'ready' || value === 'auth_required' || value === 'unavailable'
+    ? value
+    : 'unavailable';
+}
+
+/** Normalizes an untrusted `availability` to the closed set, defaulting to
+ * `unavailable` so a malformed row can never present as sendable. */
+export function asHarnessAvailability(value: unknown): HarnessModelAvailability {
+  return value === 'available' ? 'available' : 'unavailable';
+}
+
+/** A tuple is valid only when all three ids are non-empty strings — a partial
+ * or malformed selection is rejected (null) rather than half-adopted. */
+export function asHarnessSelection(value: unknown): HarnessSelection | null {
+  if (!isRecord(value)) return null;
+  const harnessId = value.harnessId;
+  const providerId = value.providerId;
+  const modelId = value.modelId;
+  if (
+    typeof harnessId !== 'string' ||
+    typeof providerId !== 'string' ||
+    typeof modelId !== 'string' ||
+    harnessId.length === 0 ||
+    providerId.length === 0 ||
+    modelId.length === 0
+  ) {
+    return null;
+  }
+  return { harnessId, providerId, modelId };
+}
+
+/** Normalizes an untrusted array into typed harness summaries, dropping any row
+ * without a usable id. */
+export function asHarnessSummaries(value: unknown): HarnessSummary[] {
+  if (!Array.isArray(value)) return [];
+  const out: HarnessSummary[] = [];
+  for (const item of value) {
+    if (!isRecord(item)) continue;
+    const id = asString(item.id);
+    if (id.length === 0) continue;
+    out.push({
+      id,
+      displayName: asNonEmptyString(item.displayName, id),
+      capabilities: asStringArray(item.capabilities),
+    });
+  }
+  return out;
+}
+
+function asHarnessCatalogEntry(value: unknown): HarnessCatalogEntry | null {
+  const selection = asHarnessSelection(value);
+  if (selection === null || !isRecord(value)) return null;
+  return {
+    ...selection,
+    displayName: asNonEmptyString(value.displayName, selection.modelId),
+    reasoningCapability: value.reasoningCapability === true,
+    inputTypes: asStringArray(value.inputTypes),
+    authState: asHarnessAuthState(value.authState),
+    availability: asHarnessAvailability(value.availability),
+  };
+}
+
+/** Normalizes an untrusted catalog body into the typed client catalog. The
+ * caller supplies `harnessId` (from the request path) so the returned catalog
+ * is scoped to the harness that was actually requested, never a body-echoed id.
+ * Malformed model rows are dropped rather than invalidating the whole catalog. */
+export function asHarnessCatalog(value: unknown, harnessId: string): HarnessCatalog {
+  const record = isRecord(value) ? value : {};
+  const rawModels = Array.isArray(record.models) ? record.models : [];
+  const models: HarnessCatalogEntry[] = [];
+  for (const row of rawModels) {
+    const entry = asHarnessCatalogEntry(row);
+    if (entry !== null) models.push(entry);
+  }
+  return {
+    harnessId,
+    version: asString(record.version),
+    fingerprint: asString(record.fingerprint),
+    models,
+  };
+}
+
+/** The single point of truth for what counts as a valid conversation ID
+ * anywhere a scoped server event may adopt one into state — a non-empty
+ * string, nothing else. Every site that establishes or compares
+ * `state.conversationId` against a raw socket payload must route through
+ * this guard so a malformed first frame (null/object/number/empty string)
+ * can never be adopted verbatim. */
+export function asConversationId(value: unknown): string | null {
+  return typeof value === 'string' && value.length > 0 ? value : null;
+}

apps/web/src/spa/chat/session-panel.tsx

@@ -0,0 +1,68 @@
+import type { ReactElement } from 'react';
+import type { SessionInfoPayload } from '@/lib/chat-contract';
+import { MAX_MANIFEST_ITEMS } from './limits';
+import { asString, asStringArray } from './runtime-guards';
+
+interface SessionPanelProps {
+  sessionInfo: SessionInfoPayload | null;
+  onSetThinking: (level: string) => void;
+}
+
+const THINKING_LEVEL_UNAVAILABLE = '';
+
+export function SessionPanel({
+  sessionInfo,
+  onSetThinking,
+}: SessionPanelProps): ReactElement | null {
+  if (!sessionInfo) return null;
+
+  // The reducer already caps this before storing it, but the render site
+  // defends independently — a hostile payload must never be able to force
+  // this <select> to lay out an unbounded number of options.
+  const availableThinkingLevels = asStringArray(sessionInfo.availableThinkingLevels).slice(
+    0,
+    MAX_MANIFEST_ITEMS,
+  );
+  const hasThinkingLevels = availableThinkingLevels.length > 0;
+
+  return (
+    <section
+      aria-label="Session info"
+      className="flex flex-wrap items-center gap-3 border-b px-4 py-2 text-xs"
+    >
+      <span>{asString(sessionInfo.provider, 'unknown')}</span>
+      <span>{asString(sessionInfo.modelId, 'unknown')}</span>
+      <label className="flex items-center gap-2">
+        <span>Thinking level</span>
+        <select
+          aria-label="Thinking level"
+          value={
+            hasThinkingLevels ? asString(sessionInfo.thinkingLevel) : THINKING_LEVEL_UNAVAILABLE
+          }
+          onChange={(event) => {
+            // The placeholder option is not a real, settable level — a
+            // malformed availableThinkingLevels list must never let the
+            // client emit set:thinking for it.
+            if (!hasThinkingLevels) return;
+            onSetThinking(event.target.value);
+          }}
+        >
+          {hasThinkingLevels ? (
+            availableThinkingLevels.map((level) => (
+              <option key={level} value={level}>
+                {level}
+              </option>
+            ))
+          ) : (
+            <option value={THINKING_LEVEL_UNAVAILABLE}>Thinking level unavailable</option>
+          )}
+        </select>
+      </label>
+      {sessionInfo.routingDecision ? (
+        <span title={asString(sessionInfo.routingDecision.ruleName)}>
+          {asString(sessionInfo.routingDecision.reason)}
+        </span>
+      ) : null}
+    </section>
+  );
+}

apps/web/src/spa/chat/test-support/fake-chat-socket.ts

@@ -0,0 +1,133 @@
+import { vi } from 'vitest';
+import type { ClientToServerEvents, ServerToClientEvents } from '@/lib/chat-contract';
+
+type ServerEvent = keyof ServerToClientEvents;
+type ClientEvent = keyof ClientToServerEvents;
+type ServerHandler<K extends ServerEvent> = ServerToClientEvents[K];
+type ClientPayload<K extends ClientEvent> = Parameters<ClientToServerEvents[K]>[0];
+
+export interface EmittedEvent<K extends ClientEvent = ClientEvent> {
+  event: K;
+  payload: ClientPayload<K>;
+}
+
+/** The subset of a Socket.IO `ChatSocket` that `useChatConnection` drives. */
+export interface FakeChatSocket {
+  connected: boolean;
+  /** Mirrors socket.io-client's `Socket.id`: the connection identity the server
+   * echoes in a `chat:send-capability` payload. The generation-bound send
+   * protocol accepts an advertisement only when `payload.connectionId === id`. */
+  id: string;
+  connect(): FakeChatSocket;
+  on<K extends ServerEvent>(event: K, handler: ServerHandler<K>): FakeChatSocket;
+  off<K extends ServerEvent>(event: K, handler: ServerHandler<K>): FakeChatSocket;
+  emit<K extends ClientEvent>(event: K, payload: ClientPayload<K>): FakeChatSocket;
+}
+
+/**
+ * A typed in-memory stand-in for `getSocket()`. Unlike a bare
+ * `(event: string, payload: unknown) => void` mock, every public method here is
+ * checked against the real `/chat` contract — a typo'd event name or a payload
+ * missing a required field fails to compile instead of silently no-op'ing at
+ * runtime.
+ */
+/** Socket.IO's built-in connection-state events. Not part of the app-level
+ * ServerToClientEvents contract, but real sockets always support them and
+ * `useChatConnection` registers a `disconnect` handler on the real socket. */
+type LifecycleEvent = 'connect' | 'disconnect';
+
+export function createFakeChatSocket(): {
+  socket: FakeChatSocket;
+  listeners: Map<ServerEvent, Set<(payload: never) => void>>;
+  emitted: EmittedEvent[];
+  serverEmit<K extends ServerEvent>(
+    event: K,
+    payload: Parameters<ServerToClientEvents[K]>[0],
+  ): void;
+  /** Escape hatch for malformed-payload tests: bypasses the compile-time
+   * payload contract to simulate a genuinely untrusted runtime value from the
+   * server, e.g. a `session:info` with a non-array `availableThinkingLevels`. */
+  serverEmitRaw(event: ServerEvent, payload: unknown): void;
+  /** Simulates a transient Socket.IO `disconnect` — fires any handler(s)
+   * registered via `socket.on('disconnect', ...)` without clearing any
+   * listeners, mirroring how a real reconnecting socket behaves. */
+  simulateDisconnect(): void;
+  /** Simulates socket.io-client's automatic reconnect of the *same*
+   * instance after a transient disconnect: marks the socket connected again
+   * and fires any handler(s) registered via `socket.on('connect', ...)`,
+   * without clearing or replacing any listeners. A real reconnect is assigned
+   * a fresh `Socket.id`; pass `nextId` to model that new connection identity
+   * (defaults to the current id so existing callers are unaffected). */
+  simulateReconnect(nextId?: string): void;
+} {
+  const listeners = new Map<ServerEvent, Set<(payload: never) => void>>();
+  const emitted: EmittedEvent[] = [];
+
+  // Internal storage is intentionally keyed loosely (the per-event handler shape
+  // varies by K, which a single Map can't express); the generic signatures on the
+  // exported `socket`/`serverEmit` above and below are what keep test call sites
+  // type-checked against ServerToClientEvents/ClientToServerEvents.
+  const socket = {
+    connected: false,
+    id: 'socket-a',
+    connect: vi.fn(function connect(this: void) {
+      socket.connected = true;
+      return socket;
+    }),
+    on: vi.fn(function on(this: void, event: ServerEvent, handler: (payload: never) => void) {
+      if (!listeners.has(event)) listeners.set(event, new Set());
+      listeners.get(event)?.add(handler);
+      return socket;
+    }),
+    off: vi.fn(function off(this: void, event: ServerEvent, handler: (payload: never) => void) {
+      listeners.get(event)?.delete(handler);
+      return socket;
+    }),
+    emit: vi.fn(function emit(this: void, event: ClientEvent, payload: unknown) {
+      emitted.push({ event, payload } as EmittedEvent);
+      return socket;
+    }),
+  } as unknown as FakeChatSocket;
+
+  function serverEmit<K extends ServerEvent>(
+    event: K,
+    payload: Parameters<ServerToClientEvents[K]>[0],
+  ): void {
+    for (const handler of listeners.get(event) ?? []) {
+      (handler as (payload: Parameters<ServerToClientEvents[K]>[0]) => void)(payload);
+    }
+  }
+
+  function serverEmitRaw(event: ServerEvent, payload: unknown): void {
+    for (const handler of listeners.get(event) ?? []) {
+      (handler as (payload: unknown) => void)(payload);
+    }
+  }
+
+  function simulateDisconnect(): void {
+    socket.connected = false;
+    const lifecycleKey = 'disconnect' satisfies LifecycleEvent as unknown as ServerEvent;
+    for (const handler of listeners.get(lifecycleKey) ?? []) {
+      (handler as () => void)();
+    }
+  }
+
+  function simulateReconnect(nextId: string = socket.id): void {
+    socket.connected = true;
+    socket.id = nextId;
+    const lifecycleKey = 'connect' satisfies LifecycleEvent as unknown as ServerEvent;
+    for (const handler of listeners.get(lifecycleKey) ?? []) {
+      (handler as () => void)();
+    }
+  }
+
+  return {
+    socket,
+    listeners,
+    emitted,
+    serverEmit,
+    serverEmitRaw,
+    simulateDisconnect,
+    simulateReconnect,
+  };
+}

apps/web/src/spa/chat/tool-call-list.spec.tsx

@@ -0,0 +1,63 @@
+import { act } from 'react';
+import { createRoot, type Root } from 'react-dom/client';
+import { afterAll, afterEach, beforeAll, describe, expect, it, vi } from 'vitest';
+import { ToolCallList } from './tool-call-list';
+
+beforeAll(() => {
+  Object.defineProperty(globalThis, 'IS_REACT_ACT_ENVIRONMENT', {
+    configurable: true,
+    value: true,
+  });
+});
+
+afterAll(() => {
+  Reflect.deleteProperty(globalThis, 'IS_REACT_ACT_ENVIRONMENT');
+});
+
+let root: Root | null;
+let container: HTMLElement | null;
+
+async function render(node: Parameters<Root['render']>[0]): Promise<void> {
+  container = document.createElement('div');
+  document.body.append(container);
+  root = createRoot(container);
+  await act(async () => {
+    root?.render(node);
+  });
+}
+
+afterEach(async () => {
+  await act(async () => {
+    root?.unmount();
+  });
+  document.body.replaceChildren();
+  root = null;
+  container = null;
+});
+
+describe('ToolCallList', () => {
+  it('renders two entries independently, without a duplicate-key warning, when a valid toolCallId is shared', async () => {
+    const consoleError = vi.spyOn(console, 'error').mockImplementation(() => {});
+
+    await render(
+      <ToolCallList
+        tools={[
+          { toolCallId: 'dup', toolName: 'search', status: 'success' },
+          { toolCallId: 'dup', toolName: 'search', status: 'running' },
+        ]}
+      />,
+    );
+
+    const items = [...(container?.querySelectorAll('li') ?? [])];
+    expect(items).toHaveLength(2);
+    expect(items[0]?.textContent).toContain('success');
+    expect(items[1]?.textContent).toContain('running');
+
+    const duplicateKeyWarning = consoleError.mock.calls.some((args) =>
+      args.some((arg) => typeof arg === 'string' && arg.includes('same key')),
+    );
+    expect(duplicateKeyWarning).toBe(false);
+
+    consoleError.mockRestore();
+  });
+});

apps/web/src/spa/chat/tool-call-list.tsx

@@ -0,0 +1,24 @@
+import type { ReactElement } from 'react';
+import type { ToolCallState } from './use-chat-connection';
+
+export function ToolCallList({ tools }: { tools: ToolCallState[] }): ReactElement | null {
+  if (tools.length === 0) return null;
+
+  return (
+    <ul aria-label="Tool calls" className="flex flex-col gap-1 px-4 pb-2 text-xs">
+      {tools.map((tool, index) => (
+        <li
+          // A valid server-controlled toolCallId can legitimately repeat
+          // (e.g. two tool:start events sharing one id) — keying on it alone
+          // would give React two identical keys. Pairing it with its
+          // (stable, append-only) render index keeps every key unique.
+          key={`${tool.toolCallId}-${index}`}
+          role={tool.status === 'error' || tool.status === 'anomaly' ? 'alert' : 'status'}
+        >
+          {tool.toolName} —{' '}
+          {tool.status === 'anomaly' ? 'unexpected end (unknown tool call)' : tool.status}
+        </li>
+      ))}
+    </ul>
+  );
+}

apps/web/src/spa/chat/use-harness-selection.spec.tsx

@@ -0,0 +1,450 @@
+import { act, type ReactElement } from 'react';
+import { createRoot, type Root } from 'react-dom/client';
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, it, vi } from 'vitest';
+import { useHarnessSelection, type HarnessSelectionValue } from './use-harness-selection';
+
+function json(body: unknown, status = 200): Response {
+  return new Response(JSON.stringify(body), {
+    status,
+    headers: { 'Content-Type': 'application/json' },
+  });
+}
+
+interface Scenario {
+  harnesses?: unknown;
+  catalog?: { body: unknown; status?: number };
+  selection?: unknown;
+  /** When set, the PUT resolves only when this is called (for race tests). */
+  deferPut?: boolean;
+}
+
+interface Deferred<T> {
+  promise: Promise<T>;
+  resolve: (value: T) => void;
+}
+
+function defer<T>(): Deferred<T> {
+  let resolve!: (value: T) => void;
+  const promise = new Promise<T>((r) => {
+    resolve = r;
+  });
+  return { promise, resolve };
+}
+
+let putBodies: unknown[] = [];
+let putDeferred: Deferred<Response> | null = null;
+
+function installFetch(scenario: Scenario): ReturnType<typeof vi.fn> {
+  putBodies = [];
+  putDeferred = scenario.deferPut ? defer<Response>() : null;
+  const fetchMock = vi.fn(async (input: unknown, init?: RequestInit) => {
+    const url = String(input);
+    const method = String(init?.method ?? 'GET').toUpperCase();
+    if (url === '/api/harnesses') return json(scenario.harnesses ?? []);
+    if (url.startsWith('/api/harnesses/') && url.endsWith('/catalog')) {
+      const spec = scenario.catalog ?? {
+        body: { harnessId: 'pi', version: '1', fingerprint: 'f', models: [] },
+      };
+      return json(spec.body, spec.status ?? 200);
+    }
+    if (url === '/api/chat/preferences/selection' && method === 'GET') {
+      return json({ selection: scenario.selection ?? null });
+    }
+    if (url === '/api/chat/preferences/selection' && method === 'PUT') {
+      putBodies.push(JSON.parse(String(init?.body)));
+      const ok = json({ selection: JSON.parse(String(init?.body)) });
+      if (putDeferred) return putDeferred.promise;
+      return ok;
+    }
+    return new Response('not found', { status: 404 });
+  });
+  vi.stubGlobal('fetch', fetchMock);
+  return fetchMock;
+}
+
+let latest: HarnessSelectionValue | null = null;
+
+function Probe(): ReactElement | null {
+  latest = useHarnessSelection();
+  return null;
+}
+
+let root: Root | null;
+let container: HTMLElement;
+
+beforeAll(() => {
+  Object.defineProperty(globalThis, 'IS_REACT_ACT_ENVIRONMENT', {
+    configurable: true,
+    value: true,
+  });
+});
+
+afterAll(() => {
+  Reflect.deleteProperty(globalThis, 'IS_REACT_ACT_ENVIRONMENT');
+});
+
+beforeEach(() => {
+  latest = null;
+  container = document.createElement('div');
+  document.body.append(container);
+  root = createRoot(container);
+});
+
+afterEach(async () => {
+  await act(async () => {
+    root?.unmount();
+  });
+  document.body.replaceChildren();
+  vi.unstubAllGlobals();
+});
+
+async function mount(): Promise<void> {
+  await act(async () => {
+    root?.render(<Probe />);
+  });
+  await flush();
+}
+
+async function flush(times = 5): Promise<void> {
+  for (let i = 0; i < times; i += 1) {
+    await act(async () => {
+      await Promise.resolve();
+    });
+  }
+}
+
+function value(): HarnessSelectionValue {
+  if (!latest) throw new Error('hook value not captured');
+  return latest;
+}
+
+const PI_CATALOG = {
+  harnessId: 'pi',
+  version: '2026-08-11',
+  fingerprint: 'fp',
+  models: [
+    {
+      harnessId: 'pi',
+      providerId: 'openai',
+      modelId: 'gpt-5',
+      displayName: 'GPT-5',
+      reasoningCapability: true,
+      inputTypes: ['text'],
+      authState: 'ready',
+      availability: 'available',
+    },
+    {
+      harnessId: 'pi',
+      providerId: 'anthropic',
+      modelId: 'claude',
+      displayName: 'Claude',
+      reasoningCapability: true,
+      inputTypes: ['text'],
+      authState: 'ready',
+      availability: 'available',
+    },
+  ],
+};
+
+describe('useHarnessSelection', () => {
+  it('loads harnesses and, once a harness is chosen, the model options come only from its catalog', async () => {
+    installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: { body: PI_CATALOG },
+      selection: null,
+    });
+    await mount();
+
+    expect(value().harnesses).toEqual([{ id: 'pi', displayName: 'Pi', capabilities: [] }]);
+    expect(value().catalog).toBeNull();
+
+    await act(async () => {
+      value().selectHarness('pi');
+    });
+    await flush();
+
+    expect(value().catalog?.harnessId).toBe('pi');
+    expect(value().catalog?.models.map((m) => m.modelId)).toEqual(['gpt-5', 'claude']);
+  });
+
+  it('does not auto-select any catalog row when there is no persisted selection (no first-row fallback)', async () => {
+    const fetchMock = installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: { body: PI_CATALOG },
+      selection: null,
+    });
+    await mount();
+    await act(async () => {
+      value().selectHarness('pi');
+    });
+    await flush();
+
+    expect(value().modelId).toBe('');
+    expect(value().persistedSelection).toBeNull();
+    expect(value().canSend).toBe(false);
+    // Nothing was persisted — no PUT fired for an unset selection.
+    const putCalls = fetchMock.mock.calls.filter(
+      (c) => String((c[1] as RequestInit)?.method).toUpperCase() === 'PUT',
+    );
+    expect(putCalls).toHaveLength(0);
+  });
+
+  it('persists the structured tuple and only enables send AFTER the PUT resolves (no race ahead of persistence)', async () => {
+    installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: { body: PI_CATALOG },
+      selection: null,
+      deferPut: true,
+    });
+    await mount();
+    await act(async () => {
+      value().selectHarness('pi');
+    });
+    await flush();
+    await act(async () => {
+      value().selectProvider('openai');
+    });
+    await act(async () => {
+      value().selectModel('openai', 'gpt-5');
+    });
+    await flush();
+
+    // PUT is in flight (deferred) — send MUST NOT be enabled yet.
+    expect(value().canSend).toBe(false);
+
+    await act(async () => {
+      putDeferred?.resolve(
+        json({ selection: { harnessId: 'pi', providerId: 'openai', modelId: 'gpt-5' } }),
+      );
+    });
+    await flush();
+
+    expect(putBodies).toContainEqual({ harnessId: 'pi', providerId: 'openai', modelId: 'gpt-5' });
+    expect(value().persistedSelection).toEqual({
+      harnessId: 'pi',
+      providerId: 'openai',
+      modelId: 'gpt-5',
+    });
+    expect(value().canSend).toBe(true);
+    // Task Five: the composer sends the nested `persistedSelection` tuple directly.
+    // The Task-Four compat flat `projection` ({provider, modelId}) is removed — the
+    // harnessId must never be dropped on the way to the wire.
+    expect('projection' in value()).toBe(false);
+  });
+
+  it('keeps a stale/unavailable persisted selection visibly displayed rather than silently dropping it', async () => {
+    installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: { body: PI_CATALOG },
+      selection: { harnessId: 'pi', providerId: 'openai', modelId: 'retired-model' },
+    });
+    await mount();
+
+    // The persisted tuple is displayed even though its model is gone from the catalog.
+    expect(value().persistedSelection).toEqual({
+      harnessId: 'pi',
+      providerId: 'openai',
+      modelId: 'retired-model',
+    });
+    expect(value().modelId).toBe('retired-model');
+    expect(value().isStale).toBe(true);
+    // A stale model is not a valid catalog option, so send stays disabled.
+    expect(value().canSend).toBe(false);
+  });
+
+  it('disables send for an empty catalog (no viable model) and never fabricates one', async () => {
+    installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: { body: { harnessId: 'pi', version: '1', fingerprint: 'f', models: [] } },
+      selection: null,
+    });
+    await mount();
+    await act(async () => {
+      value().selectHarness('pi');
+    });
+    await flush();
+
+    expect(value().catalog?.models ?? []).toHaveLength(0);
+    expect(value().canSend).toBe(false);
+  });
+
+  it('marks the catalog unavailable and disables send when the catalog request 404s', async () => {
+    installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: {
+        body: { code: 'adapter_unavailable', message: 'x', harnessId: 'pi' },
+        status: 404,
+      },
+      selection: null,
+    });
+    await mount();
+    await act(async () => {
+      value().selectHarness('pi');
+    });
+    await flush();
+
+    expect(value().catalogUnavailable).toBe(true);
+    expect(value().canSend).toBe(false);
+  });
+
+  it('on a 422 persist, keeps the requested tuple visible, surfaces a typed error, and leaves send disabled', async () => {
+    installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: {
+        body: {
+          ...PI_CATALOG,
+          models: [{ ...PI_CATALOG.models[0], availability: 'unavailable' }],
+        },
+      },
+      selection: null,
+    });
+    // Override PUT to 422.
+    const fetchMock = vi.fn(async (input: unknown, init?: RequestInit) => {
+      const url = String(input);
+      const method = String(init?.method ?? 'GET').toUpperCase();
+      if (url === '/api/harnesses')
+        return json([{ id: 'pi', displayName: 'Pi', capabilities: [] }]);
+      if (url.endsWith('/catalog')) return json(PI_CATALOG);
+      if (url === '/api/chat/preferences/selection' && method === 'GET')
+        return json({ selection: null });
+      if (url === '/api/chat/preferences/selection' && method === 'PUT') {
+        return json(
+          {
+            code: 'model_unavailable',
+            message: 'nope',
+            selection: { harnessId: 'a', providerId: 'b', modelId: 'c' },
+          },
+          422,
+        );
+      }
+      return new Response('nf', { status: 404 });
+    });
+    vi.stubGlobal('fetch', fetchMock);
+
+    await mount();
+    await act(async () => {
+      value().selectHarness('pi');
+    });
+    await flush();
+    await act(async () => {
+      value().selectProvider('openai');
+    });
+    await act(async () => {
+      value().selectModel('openai', 'gpt-5');
+    });
+    await flush();
+
+    expect(value().modelId).toBe('gpt-5');
+    expect(value().persistError?.code).toBe('model_unavailable');
+    expect(value().persistError?.requested).toEqual({
+      harnessId: 'pi',
+      providerId: 'openai',
+      modelId: 'gpt-5',
+    });
+    expect(value().persistedSelection).toBeNull();
+    expect(value().canSend).toBe(false);
+  });
+
+  it('invalidates the model on a provider change and keeps send disabled until the new tuple persists', async () => {
+    installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: { body: PI_CATALOG },
+      selection: null,
+    });
+    await mount();
+    await act(async () => {
+      value().selectHarness('pi');
+    });
+    await flush();
+    await act(async () => {
+      value().selectProvider('openai');
+    });
+    await act(async () => {
+      value().selectModel('openai', 'gpt-5');
+    });
+    await flush();
+    // A valid provider-A tuple has persisted.
+    expect(value().canSend).toBe(true);
+    expect(value().persistedSelection).toEqual({
+      harnessId: 'pi',
+      providerId: 'openai',
+      modelId: 'gpt-5',
+    });
+
+    // Switching provider clears the model that no longer belongs to it.
+    await act(async () => {
+      value().selectProvider('anthropic');
+    });
+    expect(value().modelId).toBe('');
+    expect(value().canSend).toBe(false);
+
+    // Send stays disabled until the new exact provider-B tuple persists.
+    await act(async () => {
+      value().selectModel('anthropic', 'claude');
+    });
+    await flush();
+    expect(value().canSend).toBe(true);
+    expect(value().persistedSelection).toEqual({
+      harnessId: 'pi',
+      providerId: 'anthropic',
+      modelId: 'claude',
+    });
+    // Task Five: no compat flat projection — the nested persistedSelection is the wire tuple.
+    expect('projection' in value()).toBe(false);
+  });
+
+  it('does not enable send on a model pick until the PUT for that exact new tuple resolves', async () => {
+    installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: { body: PI_CATALOG },
+      selection: { harnessId: 'pi', providerId: 'openai', modelId: 'gpt-5' },
+      deferPut: true,
+    });
+    await mount();
+    // The persisted, in-catalog tuple is sendable after mount (no PUT needed).
+    expect(value().canSend).toBe(true);
+
+    await act(async () => {
+      value().selectProvider('anthropic');
+    });
+    expect(value().modelId).toBe('');
+    expect(value().canSend).toBe(false);
+
+    await act(async () => {
+      value().selectModel('anthropic', 'claude');
+    });
+    await flush();
+    // PUT for the new tuple is still in flight — send MUST stay disabled.
+    expect(value().canSend).toBe(false);
+
+    await act(async () => {
+      putDeferred?.resolve(
+        json({ selection: { harnessId: 'pi', providerId: 'anthropic', modelId: 'claude' } }),
+      );
+    });
+    await flush();
+    expect(value().canSend).toBe(true);
+    // Task Five: no compat flat projection — the nested persistedSelection is the wire tuple.
+    expect('projection' in value()).toBe(false);
+  });
+
+  it('never requests any /api/providers* endpoint across the whole flow', async () => {
+    const fetchMock = installFetch({
+      harnesses: [{ id: 'pi', displayName: 'Pi', capabilities: [] }],
+      catalog: { body: PI_CATALOG },
+      selection: { harnessId: 'pi', providerId: 'openai', modelId: 'gpt-5' },
+    });
+    await mount();
+    await act(async () => {
+      value().selectProvider('anthropic');
+    });
+    await act(async () => {
+      value().selectModel('anthropic', 'claude');
+    });
+    await flush();
+
+    for (const call of fetchMock.mock.calls) {
+      expect(String(call[0])).not.toContain('/api/providers');
+    }
+  });
+});

apps/web/src/spa/chat/use-harness-selection.ts

@@ -0,0 +1,208 @@
+import { useCallback, useEffect, useRef, useState } from 'react';
+import {
+  fetchCatalog,
+  fetchHarnesses,
+  fetchPersistedSelection,
+  persistSelection,
+  type SelectionErrorCode,
+} from './chat-api';
+import type { HarnessCatalog, HarnessSelection, HarnessSummary } from '@/lib/types';
+
+export interface HarnessPersistError {
+  code: SelectionErrorCode;
+  message: string;
+  /** The exact tuple the user requested — preserved so the failed selection
+   * stays visible rather than being silently dropped. */
+  requested: HarnessSelection;
+}
+
+export interface HarnessSelectionValue {
+  harnesses: HarnessSummary[];
+  catalog: HarnessCatalog | null;
+  /** True when the selected harness has no usable catalog (404/error). */
+  catalogUnavailable: boolean;
+  /** The working (displayed) selection, kept as three distinct ids. Empty
+   * strings mean "not chosen yet" — there is deliberately no first-row default. */
+  harnessId: string;
+  providerId: string;
+  modelId: string;
+  /** The last tuple confirmed persisted by the server, or null. */
+  persistedSelection: HarnessSelection | null;
+  /** True when a persisted selection references a model no longer present as an
+   * available catalog entry — it stays visibly displayed rather than dropped. */
+  isStale: boolean;
+  /** True ONLY once a full tuple has been confirmed persisted AND it is a
+   * currently-available catalog entry. Send stays disabled otherwise, so a send
+   * can never race ahead of successful persistence. */
+  canSend: boolean;
+  persistError: HarnessPersistError | null;
+  selectHarness: (harnessId: string) => void;
+  selectProvider: (providerId: string) => void;
+  /** Persist the EXACT catalog row's `{providerId, modelId}` — the caller
+   * resolves the composite option identity to the real entry and passes both
+   * ids, so a bare model id is never combined with ambient provider state. */
+  selectModel: (providerId: string, modelId: string) => void;
+}
+
+/** A tuple is a currently-usable catalog option only when the catalog holds a
+ * matching, available entry — the single gate that keeps a stale/unavailable
+ * model from ever counting as sendable. */
+function isAvailableInCatalog(
+  selection: HarnessSelection | null,
+  catalog: HarnessCatalog | null,
+): boolean {
+  if (selection === null || catalog === null) return false;
+  return catalog.models.some(
+    (model) =>
+      model.providerId === selection.providerId &&
+      model.modelId === selection.modelId &&
+      model.availability === 'available',
+  );
+}
+
+function tuplesEqual(a: HarnessSelection | null, b: HarnessSelection | null): boolean {
+  if (a === null || b === null) return a === b;
+  return a.harnessId === b.harnessId && a.providerId === b.providerId && a.modelId === b.modelId;
+}
+
+/**
+ * Owns the harness/catalog/selection state for the chat composer: loads the
+ * harness list and any persisted tuple on mount, loads a harness's catalog when
+ * chosen, and PUT-persists the full `{harnessId, providerId, modelId}` tuple
+ * when a model is picked. It never auto-selects a catalog row, keeps a
+ * stale/unavailable persisted tuple visible, and only reports `canSend` true
+ * once a full tuple has actually persisted as an available catalog entry.
+ */
+export function useHarnessSelection(): HarnessSelectionValue {
+  const [harnesses, setHarnesses] = useState<HarnessSummary[]>([]);
+  const [catalog, setCatalog] = useState<HarnessCatalog | null>(null);
+  const [catalogUnavailable, setCatalogUnavailable] = useState(false);
+  const [harnessId, setHarnessId] = useState('');
+  const [providerId, setProviderId] = useState('');
+  const [modelId, setModelId] = useState('');
+  const [persistedSelection, setPersistedSelection] = useState<HarnessSelection | null>(null);
+  const [persistError, setPersistError] = useState<HarnessPersistError | null>(null);
+
+  // Monotonic request ids so a slow in-flight catalog/persist response can never
+  // overwrite the result of a newer request the user has since triggered.
+  const catalogRequestRef = useRef(0);
+  const persistRequestRef = useRef(0);
+
+  const loadCatalog = useCallback(async (id: string): Promise<void> => {
+    const requestId = catalogRequestRef.current + 1;
+    catalogRequestRef.current = requestId;
+    setCatalog(null);
+    setCatalogUnavailable(false);
+    const result = await fetchCatalog(id);
+    if (catalogRequestRef.current !== requestId) return;
+    if (result.ok) {
+      setCatalog(result.catalog);
+      setCatalogUnavailable(false);
+    } else {
+      setCatalog(null);
+      setCatalogUnavailable(true);
+    }
+  }, []);
+
+  useEffect(() => {
+    let active = true;
+    void (async (): Promise<void> => {
+      const [list, persisted] = await Promise.all([fetchHarnesses(), fetchPersistedSelection()]);
+      if (!active) return;
+      setHarnesses(list);
+      if (persisted !== null) {
+        // Adopt the persisted tuple as the displayed selection and load its
+        // catalog. If the model has since been retired, it still shows (stale).
+        setHarnessId(persisted.harnessId);
+        setProviderId(persisted.providerId);
+        setModelId(persisted.modelId);
+        setPersistedSelection(persisted);
+        await loadCatalog(persisted.harnessId);
+      }
+      // No persisted selection → nothing is auto-selected; the user must choose.
+    })();
+    return () => {
+      active = false;
+    };
+  }, [loadCatalog]);
+
+  const selectHarness = useCallback(
+    (id: string): void => {
+      setHarnessId(id);
+      // Changing harness invalidates the provider/model draft — never carry a
+      // model across harnesses.
+      setProviderId('');
+      setModelId('');
+      setPersistError(null);
+      void loadCatalog(id);
+    },
+    [loadCatalog],
+  );
+
+  const selectProvider = useCallback((id: string): void => {
+    setProviderId(id);
+    // A new provider invalidates the chosen model — no cross-provider carryover.
+    setModelId('');
+    setPersistError(null);
+  }, []);
+
+  const selectModel = useCallback(
+    (selectedProviderId: string, selectedModelId: string): void => {
+      // Bind the model to the EXACT catalog row's provider — never to ambient
+      // provider state — so two providers exposing the same modelId can never
+      // collide or mis-resolve. Keep the displayed provider consistent with the
+      // resolved row.
+      setProviderId(selectedProviderId);
+      setModelId(selectedModelId);
+      setPersistError(null);
+      const requested: HarnessSelection = {
+        harnessId,
+        providerId: selectedProviderId,
+        modelId: selectedModelId,
+      };
+      const requestId = persistRequestRef.current + 1;
+      persistRequestRef.current = requestId;
+      void (async (): Promise<void> => {
+        const result = await persistSelection(requested);
+        if (persistRequestRef.current !== requestId) return;
+        if (result.ok) {
+          setPersistedSelection(result.selection);
+          setPersistError(null);
+        } else {
+          // Leave persistedSelection unchanged (send stays disabled) and surface
+          // the typed error carrying the exact requested tuple.
+          setPersistError({
+            code: result.code,
+            message: result.message,
+            requested: result.requested,
+          });
+        }
+      })();
+    },
+    [harnessId],
+  );
+
+  const draft: HarnessSelection = { harnessId, providerId, modelId };
+  const isStale = persistedSelection !== null && !isAvailableInCatalog(persistedSelection, catalog);
+  const canSend =
+    persistedSelection !== null &&
+    !catalogUnavailable &&
+    tuplesEqual(draft, persistedSelection) &&
+    isAvailableInCatalog(persistedSelection, catalog);
+
+  return {
+    harnesses,
+    catalog,
+    catalogUnavailable,
+    harnessId,
+    providerId,
+    modelId,
+    persistedSelection,
+    isStale,
+    canSend,
+    persistError,
+    selectHarness,
+    selectProvider,
+    selectModel,
+  };
+}

apps/web/src/spa/guards.spec.tsx

@@ -0,0 +1,135 @@
+import { act } from 'react';
+import { createRoot, type Root } from 'react-dom/client';
+import { createMemoryRouter, RouterProvider, type RouteObject } from 'react-router-dom';
+import { afterAll, afterEach, beforeAll, describe, expect, it, vi } from 'vitest';
+
+const { useSessionMock } = vi.hoisted(() => ({
+  useSessionMock: vi.fn(),
+}));
+
+vi.mock('@/lib/auth-client', () => ({
+  useSession: useSessionMock,
+}));
+
+import { AuthGuard, GuestGuard } from './guards';
+
+interface RenderedRouter {
+  container: HTMLDivElement;
+  router: ReturnType<typeof createMemoryRouter>;
+}
+
+const mountedRoots: Root[] = [];
+
+beforeAll(() => {
+  Object.defineProperty(globalThis, 'IS_REACT_ACT_ENVIRONMENT', {
+    configurable: true,
+    value: true,
+  });
+});
+
+afterAll(() => {
+  Reflect.deleteProperty(globalThis, 'IS_REACT_ACT_ENVIRONMENT');
+});
+
+async function renderRouter(
+  routeObjects: RouteObject[],
+  initialEntry: string,
+): Promise<RenderedRouter> {
+  const container = document.createElement('div');
+  document.body.append(container);
+  const router = createMemoryRouter(routeObjects, { initialEntries: [initialEntry] });
+  const root = createRoot(container);
+  mountedRoots.push(root);
+
+  await act(async () => {
+    root.render(<RouterProvider router={router} />);
+  });
+
+  return { container, router };
+}
+
+afterEach(async () => {
+  for (const root of mountedRoots.splice(0)) {
+    await act(async () => {
+      root.unmount();
+    });
+  }
+  document.body.replaceChildren();
+  useSessionMock.mockReset();
+});
+
+const guestRoutes: RouteObject[] = [
+  {
+    path: '/login',
+    element: <GuestGuard />,
+    children: [{ index: true, element: <p>Guest page</p> }],
+  },
+  { path: '/chat', element: <p>Chat page</p> },
+];
+
+const authenticatedRoutes: RouteObject[] = [
+  {
+    path: '/chat',
+    element: <AuthGuard />,
+    children: [{ index: true, element: <p>Private page</p> }],
+  },
+  { path: '/login', element: <p>Login page</p> },
+];
+
+describe('GuestGuard', () => {
+  it('renders the guest outlet while session lookup is pending', async () => {
+    useSessionMock.mockReturnValue({ data: null, isPending: true });
+
+    const view = await renderRouter(guestRoutes, '/login');
+
+    expect(view.container.textContent).toContain('Guest page');
+    expect(view.router.state.location.pathname).toBe('/login');
+  });
+
+  it('renders the guest outlet when no session exists', async () => {
+    useSessionMock.mockReturnValue({ data: null, isPending: false });
+
+    const view = await renderRouter(guestRoutes, '/login');
+
+    expect(view.container.textContent).toContain('Guest page');
+    expect(view.router.state.location.pathname).toBe('/login');
+  });
+
+  it('redirects an authenticated session to chat', async () => {
+    useSessionMock.mockReturnValue({ data: { user: { id: 'user-1' } }, isPending: false });
+
+    const view = await renderRouter(guestRoutes, '/login');
+
+    expect(view.container.textContent).toContain('Chat page');
+    expect(view.router.state.location.pathname).toBe('/chat');
+  });
+});
+
+describe('AuthGuard', () => {
+  it('renders the existing loading treatment while session lookup is pending', async () => {
+    useSessionMock.mockReturnValue({ data: null, isPending: true });
+
+    const view = await renderRouter(authenticatedRoutes, '/chat');
+
+    expect(view.container.textContent).toContain('Loading...');
+    expect(view.router.state.location.pathname).toBe('/chat');
+  });
+
+  it('redirects an unauthenticated visitor to login', async () => {
+    useSessionMock.mockReturnValue({ data: null, isPending: false });
+
+    const view = await renderRouter(authenticatedRoutes, '/chat');
+
+    expect(view.container.textContent).toContain('Login page');
+    expect(view.router.state.location.pathname).toBe('/login');
+  });
+
+  it('renders the authenticated outlet when a session exists', async () => {
+    useSessionMock.mockReturnValue({ data: { user: { id: 'user-1' } }, isPending: false });
+
+    const view = await renderRouter(authenticatedRoutes, '/chat');
+
+    expect(view.container.textContent).toContain('Private page');
+    expect(view.router.state.location.pathname).toBe('/chat');
+  });
+});