diff --git a/docs/ecosystem/ECOSYSTEM_IMPLEMENTATION_TRACKER.md b/docs/ecosystem/ECOSYSTEM_IMPLEMENTATION_TRACKER.md
index 9e85fb85..c7caf2eb 100644
--- a/docs/ecosystem/ECOSYSTEM_IMPLEMENTATION_TRACKER.md
+++ b/docs/ecosystem/ECOSYSTEM_IMPLEMENTATION_TRACKER.md
@@ -242,11 +242,12 @@ These should be resolved before claiming the ecosystem docs are fully implementa
       Status note:
   - Cowork product-native runtime projections are now implemented in cowork-service
   - `023826e` adds `GET /api/agent-runtime/sessions`, `GET /api/agent-runtime/runs`, `GET /api/agent-runtime/approvals`, and `POST /api/agent-runtime/dispatch/validate`
+  - `COMMIT_PENDING` adds `GET /api/agent-runtime/tasks` with canonical `AgentTask` projection
   - FlowMonk runtime-emitter implementation work was started, but this clone cannot currently verify it because the repo depends on a local npm registry at `http://localhost:3300` and backend dependencies are not installed
 
 ### 6.1 Remaining Direct Runtime TODOs
 
-- Cowork: add `AgentTask`/`AgentTodo` direct projections once the product exposes first-class task/todo entities beyond current task-run/session surfaces.
+- Cowork: add `AgentTodo` direct projection once the product exposes first-class todo entities.
 - Cowork: attach canonical event IDs to approval and audit trails so ActionTrail lineage can stop using fallback/null semantics.
 - FlowMonk: finish direct runtime-emitter integration once the local npm registry and backend dependencies are available again.
 - Platform-service: refine the `queued -> paused` projection fallback once run-vs-session semantics are finalized.
diff --git a/docs/ecosystem/PHASE5_AGENT_RUNTIME_CONTRACT_EXECUTION_PLAN.md b/docs/ecosystem/PHASE5_AGENT_RUNTIME_CONTRACT_EXECUTION_PLAN.md
index c37a1026..f8f3a737 100644
--- a/docs/ecosystem/PHASE5_AGENT_RUNTIME_CONTRACT_EXECUTION_PLAN.md
+++ b/docs/ecosystem/PHASE5_AGENT_RUNTIME_CONTRACT_EXECUTION_PLAN.md
@@ -68,10 +68,11 @@ Observed baseline:
 - the hosted UI supports projected session review, projected run review, and dispatch payload validation
 - cowork-service now exposes direct product-native runtime projections at:
   - `GET /api/agent-runtime/sessions`
+  - `GET /api/agent-runtime/tasks`
   - `GET /api/agent-runtime/runs`
   - `GET /api/agent-runtime/approvals`
   - `POST /api/agent-runtime/dispatch/validate`
-- Cowork projections now emit canonical `AgentSession`, `AgentRun`, and `AgentApprovalCheckpoint` objects from the product backend instead of only relying on platform projections
+- Cowork projections now emit canonical `AgentSession`, `AgentTask`, `AgentRun`, and `AgentApprovalCheckpoint` objects from the product backend instead of only relying on platform projections
 
 ---
 
@@ -96,6 +97,6 @@ Observed baseline:
 
 ## 7. Remaining Gaps
 
-- Cowork now emits shared runtime projections from cowork-service, but Rust-side canonical event IDs are still missing on approval/audit records.
+- Cowork now emits shared runtime projections from cowork-service, but Rust-side canonical event IDs are still missing on approval/audit records and `AgentTodo` still has no first-class product source.
 - FlowMonk runtime-emitter code was started locally, but verification in this clone is blocked because the repo depends on a local npm registry at `http://localhost:3300` and its backend `node_modules` are missing.
 - run-vs-session semantics for queued work still need a stricter mapping than the current projection fallback.
diff --git a/services/cowork-service/src/modules/agent-runtime/routes.test.ts b/services/cowork-service/src/modules/agent-runtime/routes.test.ts
index fc90e78c..767e37ee 100644
--- a/services/cowork-service/src/modules/agent-runtime/routes.test.ts
+++ b/services/cowork-service/src/modules/agent-runtime/routes.test.ts
@@ -100,6 +100,34 @@ describe('agent runtime routes', () => {
     });
   });
 
+  it('projects IPC tasks into shared AgentTask objects', async () => {
+    call.mockResolvedValue({
+      result: {
+        tasks: [
+          {
+            id: 'task-1',
+            goal: 'Review the repository and summarize changes',
+            status: 'pending',
+            createdAt: '2026-04-04T08:00:00.000Z',
+            updatedAt: '2026-04-04T08:10:00.000Z',
+            sessionId: 'sess-1',
+          },
+        ],
+      },
+    });
+
+    const res = await app.inject({ method: 'GET', url: '/api/agent-runtime/tasks' });
+    expect(res.statusCode).toBe(200);
+    const body = JSON.parse(res.payload);
+    expect(body.tasks[0]).toMatchObject({
+      taskId: 'task-1',
+      sessionId: 'sess-1',
+      status: 'queued',
+      title: 'Review the repository and summarize changes',
+      intent: 'Review the repository and summarize changes',
+    });
+  });
+
   it('projects approval audit records into AgentApprovalCheckpoint objects', async () => {
     mockFetch
       .mockResolvedValueOnce({
diff --git a/services/cowork-service/src/modules/agent-runtime/routes.ts b/services/cowork-service/src/modules/agent-runtime/routes.ts
index 71839893..8749474c 100644
--- a/services/cowork-service/src/modules/agent-runtime/routes.ts
+++ b/services/cowork-service/src/modules/agent-runtime/routes.ts
@@ -4,9 +4,11 @@ import {
   AgentDispatchRequestSchema,
   AgentRunSchema,
   AgentSessionSchema,
+  AgentTaskSchema,
   type AgentApprovalCheckpoint,
   type AgentRun,
   type AgentSession,
+  type AgentTask,
 } from '@bytelyst/events';
 import { BadRequestError } from '@bytelyst/errors';
 import { config } from '../../lib/config.js';
@@ -104,6 +106,43 @@ function toAgentRun(task: Record<string, unknown>, fallbackNow: string): AgentRu
   });
 }
 
+function mapTaskProjectionStatus(status: unknown): AgentTask['status'] {
+  switch (status) {
+    case 'running':
+      return 'running';
+    case 'completed':
+      return 'completed';
+    case 'failed':
+      return 'failed';
+    case 'cancelled':
+      return 'cancelled';
+    case 'pending':
+    default:
+      return 'queued';
+  }
+}
+
+function toAgentTask(task: Record<string, unknown>, fallbackNow: string): AgentTask {
+  const taskId =
+    typeof task.id === 'string' && task.id.length > 0 ? task.id : `task_${fallbackNow}`;
+  const createdAt = asIsoString(task.createdAt ?? task.startedAt, fallbackNow);
+  const updatedAt = asIsoString(task.updatedAt ?? task.completedAt ?? createdAt, createdAt);
+  const goal =
+    typeof task.goal === 'string' && task.goal.trim().length > 0 ? task.goal.trim() : 'Cowork task';
+
+  return AgentTaskSchema.parse({
+    taskId,
+    sessionId:
+      typeof task.sessionId === 'string' && task.sessionId.length > 0 ? task.sessionId : taskId,
+    title: goal.length > 120 ? `${goal.slice(0, 117)}...` : goal,
+    intent: goal,
+    status: mapTaskProjectionStatus(task.status),
+    priority: typeof task.priority === 'string' ? task.priority : null,
+    createdAt,
+    updatedAt,
+  });
+}
+
 function mapApprovalStatus(action: unknown): AgentApprovalCheckpoint['status'] {
   switch (action) {
     case 'approval_granted':
@@ -235,6 +274,24 @@ export async function agentRuntimeRoutes(app: FastifyInstance) {
     };
   });
 
+  app.get('/api/agent-runtime/tasks', async req => {
+    if (!bridge.isRunning) {
+      return { tasks: [], count: 0 };
+    }
+
+    const resp = await bridge.call('list_tasks', { auth: buildAuth(req) });
+    if (resp.error) throw new BadRequestError(resp.error.message);
+
+    const raw = (resp.result as { tasks?: Record<string, unknown>[] } | undefined)?.tasks ?? [];
+    const now = new Date().toISOString();
+    const tasks = raw.map(task => toAgentTask(task, now));
+
+    return {
+      tasks,
+      count: tasks.length,
+    };
+  });
+
   app.get('/api/agent-runtime/approvals', async req => {
     const records = await fetchApprovalRecords(req);
     const now = new Date().toISOString();