diff --git a/src/config.py b/src/config.py
index 4c966e4a..83536225 100644
--- a/src/config.py
+++ b/src/config.py
@@ -149,6 +149,23 @@ class Settings(BaseSettings):
     thinking_enabled: bool = True
     thinking_interval_seconds: int = 300  # 5 minutes between thoughts
 
+    # ── Paperclip AI — orchestration bridge ────────────────────────────
+    # URL where the Paperclip server listens.
+    # For VPS deployment behind nginx, use the public domain.
+    paperclip_url: str = "http://localhost:3100"
+    # Enable/disable the Paperclip integration.
+    paperclip_enabled: bool = False
+    # API key or auth-gate cookie for authenticating with Paperclip.
+    paperclip_api_key: str = ""
+    # Timmy's agent ID in the Paperclip org chart.
+    paperclip_agent_id: str = ""
+    # Company ID in Paperclip — required for most API calls.
+    paperclip_company_id: str = ""
+    # Timeout in seconds for Paperclip HTTP calls.
+    paperclip_timeout: int = 30
+    # How often (seconds) Timmy polls Paperclip for work (0 = disabled).
+    paperclip_poll_interval: int = 0
+
     # ── OpenFang — vendored agent runtime ─────────────────────────────
     # URL where the OpenFang sidecar listens.  Set to the Docker service
     # name when running in compose, or localhost for bare-metal dev.
diff --git a/src/dashboard/app.py b/src/dashboard/app.py
index a2e514bf..264eba62 100644
--- a/src/dashboard/app.py
+++ b/src/dashboard/app.py
@@ -39,6 +39,7 @@ from dashboard.routes.thinking import router as thinking_router
 from dashboard.routes.calm import router as calm_router
 from dashboard.routes.swarm import router as swarm_router
 from dashboard.routes.system import router as system_router
+from dashboard.routes.paperclip import router as paperclip_router
 from infrastructure.router.api import router as cascade_router
 
 # Import dedicated middleware
@@ -304,6 +305,7 @@ app.include_router(thinking_router)
 app.include_router(calm_router)
 app.include_router(swarm_router)
 app.include_router(system_router)
+app.include_router(paperclip_router)
 app.include_router(cascade_router)
 
 
diff --git a/src/dashboard/routes/paperclip.py b/src/dashboard/routes/paperclip.py
new file mode 100644
index 00000000..ddd9f926
--- /dev/null
+++ b/src/dashboard/routes/paperclip.py
@@ -0,0 +1,318 @@
+"""Paperclip AI integration routes.
+
+Timmy-as-CEO: create issues, delegate to agents, review work, manage goals.
+All business logic lives in the bridge — these routes stay thin.
+"""
+
+import logging
+from typing import Optional
+
+from fastapi import APIRouter, Request
+from fastapi.responses import JSONResponse
+
+from config import settings
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter(prefix="/api/paperclip", tags=["paperclip"])
+
+
+def _disabled_response() -> JSONResponse:
+    return JSONResponse({"enabled": False, "detail": "Paperclip integration is disabled"})
+
+
+# ── Status ───────────────────────────────────────────────────────────────────
+
+
+@router.get("/status")
+async def paperclip_status():
+    """Integration health check."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    from integrations.paperclip.bridge import bridge
+
+    status = await bridge.get_status()
+    return status.model_dump()
+
+
+# ── Issues (CEO creates & manages tickets) ───────────────────────────────────
+
+
+@router.get("/issues")
+async def list_issues(status: Optional[str] = None):
+    """List all issues in the company."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    from integrations.paperclip.bridge import bridge
+
+    issues = await bridge.client.list_issues(status=status)
+    return [i.model_dump() for i in issues]
+
+
+@router.get("/issues/{issue_id}")
+async def get_issue(issue_id: str):
+    """Get issue details with comments (CEO review)."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    from integrations.paperclip.bridge import bridge
+
+    return await bridge.review_issue(issue_id)
+
+
+@router.post("/issues")
+async def create_issue(request: Request):
+    """Create a new issue and optionally assign to an agent."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    body = await request.json()
+    title = body.get("title")
+    if not title:
+        return JSONResponse({"error": "title is required"}, status_code=400)
+
+    from integrations.paperclip.bridge import bridge
+
+    issue = await bridge.create_and_assign(
+        title=title,
+        description=body.get("description", ""),
+        assignee_id=body.get("assignee_id"),
+        priority=body.get("priority"),
+        wake=body.get("wake", True),
+    )
+
+    if not issue:
+        return JSONResponse({"error": "Failed to create issue"}, status_code=502)
+
+    return issue.model_dump()
+
+
+@router.post("/issues/{issue_id}/delegate")
+async def delegate_issue(issue_id: str, request: Request):
+    """Delegate an issue to an agent (CEO assignment)."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    body = await request.json()
+    agent_id = body.get("agent_id")
+    if not agent_id:
+        return JSONResponse({"error": "agent_id is required"}, status_code=400)
+
+    from integrations.paperclip.bridge import bridge
+
+    ok = await bridge.delegate_issue(
+        issue_id=issue_id,
+        agent_id=agent_id,
+        message=body.get("message"),
+    )
+
+    if not ok:
+        return JSONResponse({"error": "Failed to delegate issue"}, status_code=502)
+
+    return {"ok": True, "issue_id": issue_id, "agent_id": agent_id}
+
+
+@router.post("/issues/{issue_id}/close")
+async def close_issue(issue_id: str, request: Request):
+    """Close an issue (CEO sign-off)."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    body = await request.json()
+
+    from integrations.paperclip.bridge import bridge
+
+    ok = await bridge.close_issue(issue_id, comment=body.get("comment"))
+
+    if not ok:
+        return JSONResponse({"error": "Failed to close issue"}, status_code=502)
+
+    return {"ok": True, "issue_id": issue_id}
+
+
+@router.post("/issues/{issue_id}/comment")
+async def add_comment(issue_id: str, request: Request):
+    """Add a CEO comment to an issue."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    body = await request.json()
+    content = body.get("content")
+    if not content:
+        return JSONResponse({"error": "content is required"}, status_code=400)
+
+    from integrations.paperclip.bridge import bridge
+
+    comment = await bridge.client.add_comment(issue_id, f"[CEO] {content}")
+
+    if not comment:
+        return JSONResponse({"error": "Failed to add comment"}, status_code=502)
+
+    return comment.model_dump()
+
+
+# ── Agents (team management) ─────────────────────────────────────────────────
+
+
+@router.get("/agents")
+async def list_agents():
+    """List all agents in the org."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    from integrations.paperclip.bridge import bridge
+
+    agents = await bridge.get_team()
+    return [a.model_dump() for a in agents]
+
+
+@router.get("/org")
+async def org_chart():
+    """Get the organizational chart."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    from integrations.paperclip.bridge import bridge
+
+    org = await bridge.get_org_chart()
+    return org or {"error": "Could not retrieve org chart"}
+
+
+@router.post("/agents/{agent_id}/wake")
+async def wake_agent(agent_id: str, request: Request):
+    """Wake an agent to start working."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    body = await request.json()
+
+    from integrations.paperclip.bridge import bridge
+
+    result = await bridge.client.wake_agent(
+        agent_id,
+        issue_id=body.get("issue_id"),
+        message=body.get("message"),
+    )
+
+    if not result:
+        return JSONResponse({"error": "Failed to wake agent"}, status_code=502)
+
+    return result
+
+
+# ── Goals ────────────────────────────────────────────────────────────────────
+
+
+@router.get("/goals")
+async def list_goals():
+    """List company goals."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    from integrations.paperclip.bridge import bridge
+
+    goals = await bridge.list_goals()
+    return [g.model_dump() for g in goals]
+
+
+@router.post("/goals")
+async def create_goal(request: Request):
+    """Set a new company goal (CEO directive)."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    body = await request.json()
+    title = body.get("title")
+    if not title:
+        return JSONResponse({"error": "title is required"}, status_code=400)
+
+    from integrations.paperclip.bridge import bridge
+
+    goal = await bridge.set_goal(title, body.get("description", ""))
+
+    if not goal:
+        return JSONResponse({"error": "Failed to create goal"}, status_code=502)
+
+    return goal.model_dump()
+
+
+# ── Approvals ────────────────────────────────────────────────────────────────
+
+
+@router.get("/approvals")
+async def list_approvals():
+    """List pending approvals for CEO review."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    from integrations.paperclip.bridge import bridge
+
+    return await bridge.pending_approvals()
+
+
+@router.post("/approvals/{approval_id}/approve")
+async def approve(approval_id: str, request: Request):
+    """Approve an agent's action."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    body = await request.json()
+
+    from integrations.paperclip.bridge import bridge
+
+    ok = await bridge.approve(approval_id, body.get("comment", ""))
+
+    if not ok:
+        return JSONResponse({"error": "Failed to approve"}, status_code=502)
+
+    return {"ok": True, "approval_id": approval_id}
+
+
+@router.post("/approvals/{approval_id}/reject")
+async def reject(approval_id: str, request: Request):
+    """Reject an agent's action."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    body = await request.json()
+
+    from integrations.paperclip.bridge import bridge
+
+    ok = await bridge.reject(approval_id, body.get("comment", ""))
+
+    if not ok:
+        return JSONResponse({"error": "Failed to reject"}, status_code=502)
+
+    return {"ok": True, "approval_id": approval_id}
+
+
+# ── Runs (monitoring) ────────────────────────────────────────────────────────
+
+
+@router.get("/runs")
+async def list_runs():
+    """List active heartbeat runs."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    from integrations.paperclip.bridge import bridge
+
+    return await bridge.active_runs()
+
+
+@router.post("/runs/{run_id}/cancel")
+async def cancel_run(run_id: str):
+    """Cancel a running heartbeat execution."""
+    if not settings.paperclip_enabled:
+        return _disabled_response()
+
+    from integrations.paperclip.bridge import bridge
+
+    ok = await bridge.cancel_run(run_id)
+
+    if not ok:
+        return JSONResponse({"error": "Failed to cancel run"}, status_code=502)
+
+    return {"ok": True, "run_id": run_id}
diff --git a/src/integrations/paperclip/__init__.py b/src/integrations/paperclip/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/src/integrations/paperclip/bridge.py b/src/integrations/paperclip/bridge.py
new file mode 100644
index 00000000..508fc9ed
--- /dev/null
+++ b/src/integrations/paperclip/bridge.py
@@ -0,0 +1,195 @@
+"""Paperclip bridge — CEO-level orchestration logic.
+
+Timmy acts as the CEO: reviews issues, delegates to agents, tracks goals,
+and approves/rejects work.  All business logic lives here; routes stay thin.
+"""
+
+from __future__ import annotations
+
+import logging
+from typing import Any, Dict, List, Optional
+
+from config import settings
+from integrations.paperclip.client import PaperclipClient, paperclip
+from integrations.paperclip.models import (
+    CreateIssueRequest,
+    PaperclipAgent,
+    PaperclipGoal,
+    PaperclipIssue,
+    PaperclipStatusResponse,
+    UpdateIssueRequest,
+)
+
+logger = logging.getLogger(__name__)
+
+
+class PaperclipBridge:
+    """Bidirectional bridge between Timmy and Paperclip.
+
+    Timmy is the CEO — he creates issues, delegates to agents via wakeup,
+    reviews results, and manages the company's goals.
+    """
+
+    def __init__(self, client: Optional[PaperclipClient] = None):
+        self.client = client or paperclip
+
+    # ── status / health ──────────────────────────────────────────────────
+
+    async def get_status(self) -> PaperclipStatusResponse:
+        """Return integration status for the dashboard."""
+        if not settings.paperclip_enabled:
+            return PaperclipStatusResponse(
+                enabled=False,
+                paperclip_url=settings.paperclip_url,
+            )
+
+        connected = await self.client.healthy()
+        agent_count = 0
+        issue_count = 0
+        error = None
+
+        if connected:
+            try:
+                agents = await self.client.list_agents()
+                agent_count = len(agents)
+                issues = await self.client.list_issues()
+                issue_count = len(issues)
+            except Exception as exc:
+                error = str(exc)
+        else:
+            error = "Cannot reach Paperclip server"
+
+        return PaperclipStatusResponse(
+            enabled=True,
+            connected=connected,
+            paperclip_url=settings.paperclip_url,
+            company_id=settings.paperclip_company_id,
+            agent_count=agent_count,
+            issue_count=issue_count,
+            error=error,
+        )
+
+    # ── CEO actions: issue management ────────────────────────────────────
+
+    async def create_and_assign(
+        self,
+        title: str,
+        description: str = "",
+        assignee_id: Optional[str] = None,
+        priority: Optional[str] = None,
+        wake: bool = True,
+    ) -> Optional[PaperclipIssue]:
+        """Create an issue and optionally assign + wake an agent.
+
+        This is the primary CEO action: decide what needs doing, create
+        the ticket, assign it to the right agent, and kick off execution.
+        """
+        req = CreateIssueRequest(
+            title=title,
+            description=description,
+            priority=priority,
+            assignee_id=assignee_id,
+        )
+        issue = await self.client.create_issue(req)
+        if not issue:
+            logger.error("Failed to create issue: %s", title)
+            return None
+
+        logger.info("Created issue %s: %s", issue.id, title)
+
+        if assignee_id and wake:
+            result = await self.client.wake_agent(assignee_id, issue_id=issue.id)
+            if result:
+                logger.info("Woke agent %s for issue %s", assignee_id, issue.id)
+            else:
+                logger.warning("Failed to wake agent %s", assignee_id)
+
+        return issue
+
+    async def delegate_issue(
+        self,
+        issue_id: str,
+        agent_id: str,
+        message: Optional[str] = None,
+    ) -> bool:
+        """Assign an existing issue to an agent and wake them."""
+        updated = await self.client.update_issue(
+            issue_id,
+            UpdateIssueRequest(assignee_id=agent_id),
+        )
+        if not updated:
+            return False
+
+        if message:
+            await self.client.add_comment(issue_id, f"[CEO] {message}")
+
+        await self.client.wake_agent(agent_id, issue_id=issue_id)
+        return True
+
+    async def review_issue(
+        self,
+        issue_id: str,
+    ) -> Dict[str, Any]:
+        """Gather all context for CEO review of an issue."""
+        issue = await self.client.get_issue(issue_id)
+        comments = await self.client.list_comments(issue_id)
+
+        return {
+            "issue": issue.model_dump() if issue else None,
+            "comments": [c.model_dump() for c in comments],
+        }
+
+    async def close_issue(self, issue_id: str, comment: Optional[str] = None) -> bool:
+        """Close an issue as the CEO."""
+        if comment:
+            await self.client.add_comment(issue_id, f"[CEO] {comment}")
+        result = await self.client.update_issue(
+            issue_id,
+            UpdateIssueRequest(status="done"),
+        )
+        return result is not None
+
+    # ── CEO actions: team management ─────────────────────────────────────
+
+    async def get_team(self) -> List[PaperclipAgent]:
+        """Get the full agent roster."""
+        return await self.client.list_agents()
+
+    async def get_org_chart(self) -> Optional[Dict[str, Any]]:
+        """Get the organizational hierarchy."""
+        return await self.client.get_org()
+
+    # ── CEO actions: goal management ─────────────────────────────────────
+
+    async def list_goals(self) -> List[PaperclipGoal]:
+        return await self.client.list_goals()
+
+    async def set_goal(self, title: str, description: str = "") -> Optional[PaperclipGoal]:
+        return await self.client.create_goal(title, description)
+
+    # ── CEO actions: approvals ───────────────────────────────────────────
+
+    async def pending_approvals(self) -> List[Dict[str, Any]]:
+        return await self.client.list_approvals()
+
+    async def approve(self, approval_id: str, comment: str = "") -> bool:
+        result = await self.client.approve(approval_id, comment)
+        return result is not None
+
+    async def reject(self, approval_id: str, comment: str = "") -> bool:
+        result = await self.client.reject(approval_id, comment)
+        return result is not None
+
+    # ── CEO actions: monitoring ──────────────────────────────────────────
+
+    async def active_runs(self) -> List[Dict[str, Any]]:
+        """Get currently running heartbeat executions."""
+        return await self.client.list_heartbeat_runs()
+
+    async def cancel_run(self, run_id: str) -> bool:
+        result = await self.client.cancel_run(run_id)
+        return result is not None
+
+
+# Module-level singleton
+bridge = PaperclipBridge()
diff --git a/src/integrations/paperclip/client.py b/src/integrations/paperclip/client.py
new file mode 100644
index 00000000..a8a33718
--- /dev/null
+++ b/src/integrations/paperclip/client.py
@@ -0,0 +1,308 @@
+"""Paperclip AI API client.
+
+Async HTTP client for communicating with a remote Paperclip server.
+All methods degrade gracefully — log the error, return a fallback, never crash.
+
+Paperclip API is mounted at ``/api`` and uses ``local_trusted`` mode on the
+VPS, so the board actor is implicit.  When the server sits behind an nginx
+auth-gate the client authenticates with Basic-auth on the first request and
+re-uses the session cookie thereafter.
+"""
+
+from __future__ import annotations
+
+import base64
+import logging
+from typing import Any, Dict, List, Optional
+
+import httpx
+
+from config import settings
+from integrations.paperclip.models import (
+    AddCommentRequest,
+    CreateIssueRequest,
+    PaperclipAgent,
+    PaperclipComment,
+    PaperclipGoal,
+    PaperclipIssue,
+    UpdateIssueRequest,
+)
+
+logger = logging.getLogger(__name__)
+
+
+class PaperclipClient:
+    """Thin async wrapper around the Paperclip REST API.
+
+    All public methods return typed results on success or ``None`` / ``[]``
+    on failure so callers never need to handle exceptions.
+    """
+
+    def __init__(
+        self,
+        base_url: Optional[str] = None,
+        api_key: Optional[str] = None,
+        timeout: int = 30,
+    ):
+        self._base_url = (base_url or settings.paperclip_url).rstrip("/")
+        self._api_key = api_key or settings.paperclip_api_key
+        self._timeout = timeout or settings.paperclip_timeout
+        self._client: Optional[httpx.AsyncClient] = None
+
+    # ── lifecycle ────────────────────────────────────────────────────────
+
+    def _get_client(self) -> httpx.AsyncClient:
+        if self._client is None or self._client.is_closed:
+            headers: Dict[str, str] = {"Accept": "application/json"}
+            if self._api_key:
+                headers["Authorization"] = f"Bearer {self._api_key}"
+            self._client = httpx.AsyncClient(
+                base_url=self._base_url,
+                headers=headers,
+                timeout=self._timeout,
+            )
+        return self._client
+
+    async def close(self) -> None:
+        if self._client and not self._client.is_closed:
+            await self._client.aclose()
+
+    # ── helpers ──────────────────────────────────────────────────────────
+
+    async def _get(self, path: str, params: Optional[Dict] = None) -> Optional[Any]:
+        try:
+            resp = await self._get_client().get(path, params=params)
+            resp.raise_for_status()
+            return resp.json()
+        except Exception as exc:
+            logger.warning("Paperclip GET %s failed: %s", path, exc)
+            return None
+
+    async def _post(self, path: str, json: Optional[Dict] = None) -> Optional[Any]:
+        try:
+            resp = await self._get_client().post(path, json=json)
+            resp.raise_for_status()
+            return resp.json()
+        except Exception as exc:
+            logger.warning("Paperclip POST %s failed: %s", path, exc)
+            return None
+
+    async def _patch(self, path: str, json: Optional[Dict] = None) -> Optional[Any]:
+        try:
+            resp = await self._get_client().patch(path, json=json)
+            resp.raise_for_status()
+            return resp.json()
+        except Exception as exc:
+            logger.warning("Paperclip PATCH %s failed: %s", path, exc)
+            return None
+
+    async def _delete(self, path: str) -> bool:
+        try:
+            resp = await self._get_client().delete(path)
+            resp.raise_for_status()
+            return True
+        except Exception as exc:
+            logger.warning("Paperclip DELETE %s failed: %s", path, exc)
+            return False
+
+    # ── health ───────────────────────────────────────────────────────────
+
+    async def healthy(self) -> bool:
+        """Quick connectivity check."""
+        data = await self._get("/api/health")
+        return data is not None
+
+    # ── companies ────────────────────────────────────────────────────────
+
+    async def list_companies(self) -> List[Dict[str, Any]]:
+        data = await self._get("/api/companies")
+        return data if isinstance(data, list) else []
+
+    # ── agents ───────────────────────────────────────────────────────────
+
+    async def list_agents(self, company_id: Optional[str] = None) -> List[PaperclipAgent]:
+        cid = company_id or settings.paperclip_company_id
+        if not cid:
+            logger.warning("paperclip_company_id not set — cannot list agents")
+            return []
+        data = await self._get(f"/api/companies/{cid}/agents")
+        if not isinstance(data, list):
+            return []
+        return [PaperclipAgent(**a) for a in data]
+
+    async def get_agent(self, agent_id: str) -> Optional[PaperclipAgent]:
+        data = await self._get(f"/api/agents/{agent_id}")
+        return PaperclipAgent(**data) if data else None
+
+    async def wake_agent(
+        self,
+        agent_id: str,
+        issue_id: Optional[str] = None,
+        message: Optional[str] = None,
+    ) -> Optional[Dict[str, Any]]:
+        """Trigger a heartbeat wake for an agent."""
+        body: Dict[str, Any] = {}
+        if issue_id:
+            body["issueId"] = issue_id
+        if message:
+            body["message"] = message
+        return await self._post(f"/api/agents/{agent_id}/wakeup", json=body)
+
+    async def get_org(self, company_id: Optional[str] = None) -> Optional[Dict[str, Any]]:
+        cid = company_id or settings.paperclip_company_id
+        if not cid:
+            return None
+        return await self._get(f"/api/companies/{cid}/org")
+
+    # ── issues (tickets) ─────────────────────────────────────────────────
+
+    async def list_issues(
+        self,
+        company_id: Optional[str] = None,
+        status: Optional[str] = None,
+    ) -> List[PaperclipIssue]:
+        cid = company_id or settings.paperclip_company_id
+        if not cid:
+            return []
+        params: Dict[str, str] = {}
+        if status:
+            params["status"] = status
+        data = await self._get(f"/api/companies/{cid}/issues", params=params)
+        if not isinstance(data, list):
+            return []
+        return [PaperclipIssue(**i) for i in data]
+
+    async def get_issue(self, issue_id: str) -> Optional[PaperclipIssue]:
+        data = await self._get(f"/api/issues/{issue_id}")
+        return PaperclipIssue(**data) if data else None
+
+    async def create_issue(
+        self,
+        req: CreateIssueRequest,
+        company_id: Optional[str] = None,
+    ) -> Optional[PaperclipIssue]:
+        cid = company_id or settings.paperclip_company_id
+        if not cid:
+            logger.warning("paperclip_company_id not set — cannot create issue")
+            return None
+        data = await self._post(
+            f"/api/companies/{cid}/issues",
+            json=req.model_dump(exclude_none=True),
+        )
+        return PaperclipIssue(**data) if data else None
+
+    async def update_issue(
+        self,
+        issue_id: str,
+        req: UpdateIssueRequest,
+    ) -> Optional[PaperclipIssue]:
+        data = await self._patch(
+            f"/api/issues/{issue_id}",
+            json=req.model_dump(exclude_none=True),
+        )
+        return PaperclipIssue(**data) if data else None
+
+    async def delete_issue(self, issue_id: str) -> bool:
+        return await self._delete(f"/api/issues/{issue_id}")
+
+    # ── issue comments ───────────────────────────────────────────────────
+
+    async def list_comments(self, issue_id: str) -> List[PaperclipComment]:
+        data = await self._get(f"/api/issues/{issue_id}/comments")
+        if not isinstance(data, list):
+            return []
+        return [PaperclipComment(**c) for c in data]
+
+    async def add_comment(
+        self,
+        issue_id: str,
+        content: str,
+    ) -> Optional[PaperclipComment]:
+        data = await self._post(
+            f"/api/issues/{issue_id}/comments",
+            json={"content": content},
+        )
+        return PaperclipComment(**data) if data else None
+
+    # ── issue workflow ───────────────────────────────────────────────────
+
+    async def checkout_issue(self, issue_id: str) -> Optional[Dict[str, Any]]:
+        """Assign an issue to Timmy (checkout)."""
+        body: Dict[str, Any] = {}
+        if settings.paperclip_agent_id:
+            body["agentId"] = settings.paperclip_agent_id
+        return await self._post(f"/api/issues/{issue_id}/checkout", json=body)
+
+    async def release_issue(self, issue_id: str) -> Optional[Dict[str, Any]]:
+        """Release a checked-out issue."""
+        return await self._post(f"/api/issues/{issue_id}/release")
+
+    # ── goals ────────────────────────────────────────────────────────────
+
+    async def list_goals(self, company_id: Optional[str] = None) -> List[PaperclipGoal]:
+        cid = company_id or settings.paperclip_company_id
+        if not cid:
+            return []
+        data = await self._get(f"/api/companies/{cid}/goals")
+        if not isinstance(data, list):
+            return []
+        return [PaperclipGoal(**g) for g in data]
+
+    async def create_goal(
+        self,
+        title: str,
+        description: str = "",
+        company_id: Optional[str] = None,
+    ) -> Optional[PaperclipGoal]:
+        cid = company_id or settings.paperclip_company_id
+        if not cid:
+            return None
+        data = await self._post(
+            f"/api/companies/{cid}/goals",
+            json={"title": title, "description": description},
+        )
+        return PaperclipGoal(**data) if data else None
+
+    # ── heartbeat runs ───────────────────────────────────────────────────
+
+    async def list_heartbeat_runs(
+        self,
+        company_id: Optional[str] = None,
+    ) -> List[Dict[str, Any]]:
+        cid = company_id or settings.paperclip_company_id
+        if not cid:
+            return []
+        data = await self._get(f"/api/companies/{cid}/heartbeat-runs")
+        return data if isinstance(data, list) else []
+
+    async def get_run_events(self, run_id: str) -> List[Dict[str, Any]]:
+        data = await self._get(f"/api/heartbeat-runs/{run_id}/events")
+        return data if isinstance(data, list) else []
+
+    async def cancel_run(self, run_id: str) -> Optional[Dict[str, Any]]:
+        return await self._post(f"/api/heartbeat-runs/{run_id}/cancel")
+
+    # ── approvals ────────────────────────────────────────────────────────
+
+    async def list_approvals(self, company_id: Optional[str] = None) -> List[Dict[str, Any]]:
+        cid = company_id or settings.paperclip_company_id
+        if not cid:
+            return []
+        data = await self._get(f"/api/companies/{cid}/approvals")
+        return data if isinstance(data, list) else []
+
+    async def approve(self, approval_id: str, comment: str = "") -> Optional[Dict[str, Any]]:
+        body: Dict[str, Any] = {}
+        if comment:
+            body["comment"] = comment
+        return await self._post(f"/api/approvals/{approval_id}/approve", json=body)
+
+    async def reject(self, approval_id: str, comment: str = "") -> Optional[Dict[str, Any]]:
+        body: Dict[str, Any] = {}
+        if comment:
+            body["comment"] = comment
+        return await self._post(f"/api/approvals/{approval_id}/reject", json=body)
+
+
+# Module-level singleton
+paperclip = PaperclipClient()
diff --git a/src/integrations/paperclip/models.py b/src/integrations/paperclip/models.py
new file mode 100644
index 00000000..c0f0c451
--- /dev/null
+++ b/src/integrations/paperclip/models.py
@@ -0,0 +1,120 @@
+"""Pydantic models for Paperclip AI API objects."""
+
+from __future__ import annotations
+
+from datetime import datetime
+from typing import Any, Dict, List, Optional
+
+from pydantic import BaseModel, Field
+
+
+# ── Inbound: Paperclip → Timmy ──────────────────────────────────────────────
+
+
+class PaperclipIssue(BaseModel):
+    """A ticket/issue in Paperclip's task system."""
+
+    id: str
+    title: str
+    description: str = ""
+    status: str = "open"
+    priority: Optional[str] = None
+    assignee_id: Optional[str] = None
+    project_id: Optional[str] = None
+    labels: List[str] = Field(default_factory=list)
+    created_at: Optional[str] = None
+    updated_at: Optional[str] = None
+
+
+class PaperclipComment(BaseModel):
+    """A comment on a Paperclip issue."""
+
+    id: str
+    issue_id: str
+    content: str
+    author: Optional[str] = None
+    created_at: Optional[str] = None
+
+
+class PaperclipAgent(BaseModel):
+    """An agent in the Paperclip org chart."""
+
+    id: str
+    name: str
+    role: str = ""
+    status: str = "active"
+    adapter_type: Optional[str] = None
+    company_id: Optional[str] = None
+
+
+class PaperclipGoal(BaseModel):
+    """A company goal in Paperclip."""
+
+    id: str
+    title: str
+    description: str = ""
+    status: str = "active"
+    company_id: Optional[str] = None
+
+
+class HeartbeatRun(BaseModel):
+    """A heartbeat execution run."""
+
+    id: str
+    agent_id: str
+    status: str
+    issue_id: Optional[str] = None
+    started_at: Optional[str] = None
+    finished_at: Optional[str] = None
+
+
+# ── Outbound: Timmy → Paperclip ─────────────────────────────────────────────
+
+
+class CreateIssueRequest(BaseModel):
+    """Request to create a new issue in Paperclip."""
+
+    title: str
+    description: str = ""
+    priority: Optional[str] = None
+    assignee_id: Optional[str] = None
+    project_id: Optional[str] = None
+    labels: List[str] = Field(default_factory=list)
+
+
+class UpdateIssueRequest(BaseModel):
+    """Request to update an existing issue."""
+
+    title: Optional[str] = None
+    description: Optional[str] = None
+    status: Optional[str] = None
+    priority: Optional[str] = None
+    assignee_id: Optional[str] = None
+
+
+class AddCommentRequest(BaseModel):
+    """Request to add a comment to an issue."""
+
+    content: str
+
+
+class WakeAgentRequest(BaseModel):
+    """Request to wake an agent via heartbeat."""
+
+    issue_id: Optional[str] = None
+    message: Optional[str] = None
+
+
+# ── API route models ─────────────────────────────────────────────────────────
+
+
+class PaperclipStatusResponse(BaseModel):
+    """Response for GET /api/paperclip/status."""
+
+    enabled: bool
+    connected: bool = False
+    paperclip_url: str = ""
+    company_id: str = ""
+    agent_count: int = 0
+    issue_count: int = 0
+    error: Optional[str] = None
diff --git a/src/integrations/paperclip/task_runner.py b/src/integrations/paperclip/task_runner.py
new file mode 100644
index 00000000..c4a9d3df
--- /dev/null
+++ b/src/integrations/paperclip/task_runner.py
@@ -0,0 +1,214 @@
+"""Paperclip task runner — automated issue processing loop.
+
+Timmy grabs open issues assigned to him, processes each one, posts a
+completion comment, marks the issue done, and creates a recursive
+follow-up task for himself.
+
+Green-path workflow:
+1. Poll Paperclip for open issues assigned to Timmy
+2. Check out the first issue in queue
+3. Process it (delegate to orchestrator via execute_task)
+4. Post completion comment with the result
+5. Mark the issue done
+6. Create a follow-up task for himself (recursive musing)
+"""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+from typing import Any, Callable, Coroutine, Dict, List, Optional, Protocol, runtime_checkable
+
+from config import settings
+from integrations.paperclip.bridge import PaperclipBridge, bridge as default_bridge
+from integrations.paperclip.models import PaperclipIssue
+
+logger = logging.getLogger(__name__)
+
+
+@runtime_checkable
+class Orchestrator(Protocol):
+    """Anything with an ``execute_task`` matching Timmy's orchestrator."""
+
+    async def execute_task(
+        self, task_id: str, description: str, context: dict
+    ) -> Any: ...
+
+
+def _wrap_orchestrator(orch: Orchestrator) -> Callable:
+    """Adapt an orchestrator's execute_task to the process_fn signature."""
+
+    async def _process(task_id: str, description: str, context: Dict) -> str:
+        raw = await orch.execute_task(task_id, description, context)
+        # execute_task may return str or dict — normalise to str
+        if isinstance(raw, dict):
+            return raw.get("result", str(raw))
+        return str(raw)
+
+    return _process
+
+
+class TaskRunner:
+    """Autonomous task loop: grab → process → complete → follow-up.
+
+    Wire an *orchestrator* (anything with ``execute_task``) and the runner
+    pushes issues through the real agent pipe.  Falls back to a plain
+    ``process_fn`` callable or a no-op default.
+
+    The runner operates on a single cycle via ``run_once`` (testable) or
+    continuously via ``start`` with ``paperclip_poll_interval``.
+    """
+
+    def __init__(
+        self,
+        bridge: Optional[PaperclipBridge] = None,
+        orchestrator: Optional[Orchestrator] = None,
+        process_fn: Optional[Callable[[str, str, Dict], Coroutine[Any, Any, str]]] = None,
+    ):
+        self.bridge = bridge or default_bridge
+        self.orchestrator = orchestrator
+
+        # Priority: explicit process_fn > orchestrator wrapper > default
+        if process_fn:
+            self._process_fn = process_fn
+        elif orchestrator:
+            self._process_fn = _wrap_orchestrator(orchestrator)
+        else:
+            self._process_fn = None
+
+        self._running = False
+
+    # ── single cycle ──────────────────────────────────────────────────
+
+    async def grab_next_task(self) -> Optional[PaperclipIssue]:
+        """Grab the first open issue assigned to Timmy."""
+        agent_id = settings.paperclip_agent_id
+        if not agent_id:
+            logger.warning("paperclip_agent_id not set — cannot grab tasks")
+            return None
+
+        issues = await self.bridge.client.list_issues(status="open")
+        # Filter to issues assigned to Timmy, take the first one
+        for issue in issues:
+            if issue.assignee_id == agent_id:
+                return issue
+
+        return None
+
+    async def process_task(self, issue: PaperclipIssue) -> str:
+        """Process an issue: check out, run through the orchestrator, return result."""
+        # Check out the issue so others know we're working on it
+        await self.bridge.client.checkout_issue(issue.id)
+
+        context = {
+            "issue_id": issue.id,
+            "title": issue.title,
+            "priority": issue.priority,
+            "labels": issue.labels,
+        }
+
+        if self._process_fn:
+            result = await self._process_fn(issue.id, issue.description or issue.title, context)
+        else:
+            result = f"Processed task: {issue.title}"
+
+        return result
+
+    async def complete_task(self, issue: PaperclipIssue, result: str) -> bool:
+        """Post completion comment and mark issue done."""
+        # Post the result as a comment
+        await self.bridge.client.add_comment(
+            issue.id,
+            f"[Timmy] Task completed.\n\n{result}",
+        )
+
+        # Mark the issue as done
+        return await self.bridge.close_issue(issue.id, comment=None)
+
+    async def create_follow_up(self, original: PaperclipIssue, result: str) -> Optional[PaperclipIssue]:
+        """Create a recursive follow-up task for Timmy.
+
+        Timmy muses about task automation and writes a follow-up issue
+        assigned to himself — the recursive self-improvement loop.
+        """
+        follow_up_title = f"Follow-up: {original.title}"
+        follow_up_description = (
+            f"Automated follow-up from completed task '{original.title}' "
+            f"(issue {original.id}).\n\n"
+            f"Previous result:\n{result}\n\n"
+            "Review the outcome and determine if further action is needed. "
+            "Muse about task automation improvements and recursive self-improvement."
+        )
+
+        return await self.bridge.create_and_assign(
+            title=follow_up_title,
+            description=follow_up_description,
+            assignee_id=settings.paperclip_agent_id,
+            priority=original.priority,
+            wake=False,  # Don't wake immediately — let the next poll pick it up
+        )
+
+    async def run_once(self) -> Optional[Dict[str, Any]]:
+        """Execute one full cycle of the green-path workflow.
+
+        Returns a summary dict on success, None if no work found.
+        """
+        # Step 1: Grab next task
+        issue = await self.grab_next_task()
+        if not issue:
+            logger.debug("No tasks in queue for Timmy")
+            return None
+
+        logger.info("Grabbed task %s: %s", issue.id, issue.title)
+
+        # Step 2: Process the task
+        result = await self.process_task(issue)
+        logger.info("Processed task %s", issue.id)
+
+        # Step 3: Complete it
+        completed = await self.complete_task(issue, result)
+        if not completed:
+            logger.warning("Failed to mark task %s as done", issue.id)
+
+        # Step 4: Create follow-up
+        follow_up = await self.create_follow_up(issue, result)
+        follow_up_id = follow_up.id if follow_up else None
+        if follow_up:
+            logger.info("Created follow-up %s for task %s", follow_up.id, issue.id)
+
+        return {
+            "original_issue_id": issue.id,
+            "original_title": issue.title,
+            "result": result,
+            "completed": completed,
+            "follow_up_issue_id": follow_up_id,
+        }
+
+    # ── continuous loop ───────────────────────────────────────────────
+
+    async def start(self) -> None:
+        """Run the task loop continuously using paperclip_poll_interval."""
+        interval = settings.paperclip_poll_interval
+        if interval <= 0:
+            logger.info("Task runner disabled (poll_interval=%d)", interval)
+            return
+
+        self._running = True
+        logger.info("Task runner started (poll every %ds)", interval)
+
+        while self._running:
+            try:
+                await self.run_once()
+            except Exception as exc:
+                logger.error("Task runner cycle failed: %s", exc)
+
+            await asyncio.sleep(interval)
+
+    def stop(self) -> None:
+        """Signal the loop to stop."""
+        self._running = False
+        logger.info("Task runner stopping")
+
+
+# Module-level singleton
+task_runner = TaskRunner()
diff --git a/tests/dashboard/test_paperclip_routes.py b/tests/dashboard/test_paperclip_routes.py
new file mode 100644
index 00000000..469c0f39
--- /dev/null
+++ b/tests/dashboard/test_paperclip_routes.py
@@ -0,0 +1,148 @@
+"""Tests for the Paperclip API routes."""
+
+from unittest.mock import AsyncMock, patch, MagicMock
+
+from integrations.paperclip.models import PaperclipIssue, PaperclipAgent, PaperclipGoal
+
+
+# ── GET /api/paperclip/status ────────────────────────────────────────────────
+
+
+def test_status_disabled(client):
+    """When paperclip_enabled is False, status returns disabled."""
+    response = client.get("/api/paperclip/status")
+    assert response.status_code == 200
+    data = response.json()
+    assert data["enabled"] is False
+
+
+def test_status_enabled(client):
+    mock_status = MagicMock()
+    mock_status.model_dump.return_value = {
+        "enabled": True,
+        "connected": True,
+        "paperclip_url": "http://vps:3100",
+        "company_id": "comp-1",
+        "agent_count": 3,
+        "issue_count": 5,
+        "error": None,
+    }
+    mock_bridge = MagicMock()
+    mock_bridge.get_status = AsyncMock(return_value=mock_status)
+    with patch("dashboard.routes.paperclip.settings") as mock_settings:
+        mock_settings.paperclip_enabled = True
+        with patch.dict("sys.modules", {}):
+            with patch("integrations.paperclip.bridge.bridge", mock_bridge):
+                response = client.get("/api/paperclip/status")
+    assert response.status_code == 200
+    assert response.json()["connected"] is True
+
+
+# ── GET /api/paperclip/issues ────────────────────────────────────────────────
+
+
+def test_list_issues_disabled(client):
+    response = client.get("/api/paperclip/issues")
+    assert response.status_code == 200
+    assert response.json()["enabled"] is False
+
+
+# ── POST /api/paperclip/issues ───────────────────────────────────────────────
+
+
+def test_create_issue_disabled(client):
+    response = client.post(
+        "/api/paperclip/issues",
+        json={"title": "Test"},
+    )
+    assert response.status_code == 200
+    assert response.json()["enabled"] is False
+
+
+def test_create_issue_missing_title(client):
+    with patch("dashboard.routes.paperclip.settings") as mock_settings:
+        mock_settings.paperclip_enabled = True
+        response = client.post(
+            "/api/paperclip/issues",
+            json={"description": "No title"},
+        )
+    assert response.status_code == 400
+    assert "title" in response.json()["error"]
+
+
+# ── POST /api/paperclip/issues/{id}/delegate ─────────────────────────────────
+
+
+def test_delegate_issue_missing_agent_id(client):
+    with patch("dashboard.routes.paperclip.settings") as mock_settings:
+        mock_settings.paperclip_enabled = True
+        response = client.post(
+            "/api/paperclip/issues/i1/delegate",
+            json={"message": "Do this"},
+        )
+    assert response.status_code == 400
+    assert "agent_id" in response.json()["error"]
+
+
+# ── POST /api/paperclip/issues/{id}/comment ──────────────────────────────────
+
+
+def test_add_comment_missing_content(client):
+    with patch("dashboard.routes.paperclip.settings") as mock_settings:
+        mock_settings.paperclip_enabled = True
+        response = client.post(
+            "/api/paperclip/issues/i1/comment",
+            json={},
+        )
+    assert response.status_code == 400
+    assert "content" in response.json()["error"]
+
+
+# ── GET /api/paperclip/agents ────────────────────────────────────────────────
+
+
+def test_list_agents_disabled(client):
+    response = client.get("/api/paperclip/agents")
+    assert response.status_code == 200
+    assert response.json()["enabled"] is False
+
+
+# ── GET /api/paperclip/goals ─────────────────────────────────────────────────
+
+
+def test_list_goals_disabled(client):
+    response = client.get("/api/paperclip/goals")
+    assert response.status_code == 200
+    assert response.json()["enabled"] is False
+
+
+# ── POST /api/paperclip/goals ────────────────────────────────────────────────
+
+
+def test_create_goal_missing_title(client):
+    with patch("dashboard.routes.paperclip.settings") as mock_settings:
+        mock_settings.paperclip_enabled = True
+        response = client.post(
+            "/api/paperclip/goals",
+            json={"description": "No title"},
+        )
+    assert response.status_code == 400
+    assert "title" in response.json()["error"]
+
+
+# ── GET /api/paperclip/approvals ─────────────────────────────────────────────
+
+
+def test_list_approvals_disabled(client):
+    response = client.get("/api/paperclip/approvals")
+    assert response.status_code == 200
+    assert response.json()["enabled"] is False
+
+
+# ── GET /api/paperclip/runs ──────────────────────────────────────────────────
+
+
+def test_list_runs_disabled(client):
+    response = client.get("/api/paperclip/runs")
+    assert response.status_code == 200
+    assert response.json()["enabled"] is False
diff --git a/tests/integrations/test_paperclip_bridge.py b/tests/integrations/test_paperclip_bridge.py
new file mode 100644
index 00000000..e1ca36cb
--- /dev/null
+++ b/tests/integrations/test_paperclip_bridge.py
@@ -0,0 +1,205 @@
+"""Tests for the Paperclip bridge (CEO orchestration logic)."""
+
+from unittest.mock import AsyncMock, patch, MagicMock
+
+import pytest
+
+from integrations.paperclip.bridge import PaperclipBridge
+from integrations.paperclip.client import PaperclipClient
+from integrations.paperclip.models import PaperclipAgent, PaperclipGoal, PaperclipIssue
+
+
+@pytest.fixture
+def mock_client():
+    client = MagicMock(spec=PaperclipClient)
+    # Make all methods async
+    client.healthy = AsyncMock(return_value=True)
+    client.list_agents = AsyncMock(return_value=[])
+    client.list_issues = AsyncMock(return_value=[])
+    client.list_goals = AsyncMock(return_value=[])
+    client.list_approvals = AsyncMock(return_value=[])
+    client.list_heartbeat_runs = AsyncMock(return_value=[])
+    client.get_issue = AsyncMock(return_value=None)
+    client.get_org = AsyncMock(return_value=None)
+    client.create_issue = AsyncMock(return_value=None)
+    client.update_issue = AsyncMock(return_value=None)
+    client.add_comment = AsyncMock(return_value=None)
+    client.wake_agent = AsyncMock(return_value=None)
+    client.create_goal = AsyncMock(return_value=None)
+    client.approve = AsyncMock(return_value=None)
+    client.reject = AsyncMock(return_value=None)
+    client.cancel_run = AsyncMock(return_value=None)
+    client.list_comments = AsyncMock(return_value=[])
+    return client
+
+
+@pytest.fixture
+def bridge(mock_client):
+    return PaperclipBridge(client=mock_client)
+
+
+# ── status ───────────────────────────────────────────────────────────────────
+
+
+async def test_status_when_disabled(bridge):
+    with patch("integrations.paperclip.bridge.settings") as mock_settings:
+        mock_settings.paperclip_enabled = False
+        mock_settings.paperclip_url = "http://localhost:3100"
+        status = await bridge.get_status()
+    assert status.enabled is False
+
+
+async def test_status_when_connected(bridge, mock_client):
+    mock_client.healthy.return_value = True
+    mock_client.list_agents.return_value = [
+        PaperclipAgent(id="a1", name="Codex"),
+    ]
+    mock_client.list_issues.return_value = [
+        PaperclipIssue(id="i1", title="Bug"),
+        PaperclipIssue(id="i2", title="Feature"),
+    ]
+
+    with patch("integrations.paperclip.bridge.settings") as mock_settings:
+        mock_settings.paperclip_enabled = True
+        mock_settings.paperclip_url = "http://vps:3100"
+        mock_settings.paperclip_company_id = "comp-1"
+        status = await bridge.get_status()
+
+    assert status.enabled is True
+    assert status.connected is True
+    assert status.agent_count == 1
+    assert status.issue_count == 2
+
+
+async def test_status_when_disconnected(bridge, mock_client):
+    mock_client.healthy.return_value = False
+
+    with patch("integrations.paperclip.bridge.settings") as mock_settings:
+        mock_settings.paperclip_enabled = True
+        mock_settings.paperclip_url = "http://vps:3100"
+        mock_settings.paperclip_company_id = "comp-1"
+        status = await bridge.get_status()
+
+    assert status.enabled is True
+    assert status.connected is False
+    assert "Cannot reach" in status.error
+
+
+# ── create and assign ────────────────────────────────────────────────────────
+
+
+async def test_create_and_assign_with_wake(bridge, mock_client):
+    issue = PaperclipIssue(id="i1", title="Deploy v2")
+    mock_client.create_issue.return_value = issue
+    mock_client.wake_agent.return_value = {"status": "queued"}
+
+    result = await bridge.create_and_assign(
+        title="Deploy v2",
+        assignee_id="agent-codex",
+        wake=True,
+    )
+
+    assert result is not None
+    assert result.id == "i1"
+    mock_client.wake_agent.assert_awaited_once_with("agent-codex", issue_id="i1")
+
+
+async def test_create_and_assign_no_wake(bridge, mock_client):
+    issue = PaperclipIssue(id="i2", title="Research task")
+    mock_client.create_issue.return_value = issue
+
+    result = await bridge.create_and_assign(
+        title="Research task",
+        assignee_id="agent-research",
+        wake=False,
+    )
+
+    assert result is not None
+    mock_client.wake_agent.assert_not_awaited()
+
+
+async def test_create_and_assign_failure(bridge, mock_client):
+    mock_client.create_issue.return_value = None
+
+    result = await bridge.create_and_assign(title="Will fail")
+    assert result is None
+
+
+# ── delegate ─────────────────────────────────────────────────────────────────
+
+
+async def test_delegate_issue(bridge, mock_client):
+    mock_client.update_issue.return_value = PaperclipIssue(id="i1", title="Task")
+    mock_client.wake_agent.return_value = {"status": "queued"}
+
+    ok = await bridge.delegate_issue("i1", "agent-codex", message="Handle this")
+    assert ok is True
+    mock_client.add_comment.assert_awaited_once()
+    mock_client.wake_agent.assert_awaited_once()
+
+
+async def test_delegate_issue_update_fails(bridge, mock_client):
+    mock_client.update_issue.return_value = None
+
+    ok = await bridge.delegate_issue("i1", "agent-codex")
+    assert ok is False
+
+
+# ── close issue ──────────────────────────────────────────────────────────────
+
+
+async def test_close_issue(bridge, mock_client):
+    mock_client.update_issue.return_value = PaperclipIssue(id="i1", title="Done")
+
+    ok = await bridge.close_issue("i1", comment="Shipped!")
+    assert ok is True
+    mock_client.add_comment.assert_awaited_once()
+
+
+# ── goals ────────────────────────────────────────────────────────────────────
+
+
+async def test_set_goal(bridge, mock_client):
+    mock_client.create_goal.return_value = PaperclipGoal(id="g1", title="Ship MVP")
+
+    goal = await bridge.set_goal("Ship MVP")
+    assert goal is not None
+    assert goal.title == "Ship MVP"
+
+
+# ── approvals ────────────────────────────────────────────────────────────────
+
+
+async def test_approve(bridge, mock_client):
+    mock_client.approve.return_value = {"status": "approved"}
+    ok = await bridge.approve("ap1")
+    assert ok is True
+
+
+async def test_reject(bridge, mock_client):
+    mock_client.reject.return_value = {"status": "rejected"}
+    ok = await bridge.reject("ap1", comment="Needs work")
+    assert ok is True
+
+
+async def test_approve_failure(bridge, mock_client):
+    mock_client.approve.return_value = None
+    ok = await bridge.approve("ap1")
+    assert ok is False
+
+
+# ── runs ─────────────────────────────────────────────────────────────────────
+
+
+async def test_active_runs(bridge, mock_client):
+    mock_client.list_heartbeat_runs.return_value = [
+        {"id": "r1", "status": "running"},
+    ]
+    runs = await bridge.active_runs()
+    assert len(runs) == 1
+
+
+async def test_cancel_run(bridge, mock_client):
+    mock_client.cancel_run.return_value = {"status": "cancelled"}
+    ok = await bridge.cancel_run("r1")
+    assert ok is True
diff --git a/tests/integrations/test_paperclip_client.py b/tests/integrations/test_paperclip_client.py
new file mode 100644
index 00000000..0bfdf82a
--- /dev/null
+++ b/tests/integrations/test_paperclip_client.py
@@ -0,0 +1,180 @@
+"""Tests for the Paperclip API client."""
+
+from unittest.mock import AsyncMock, patch
+
+import pytest
+
+from integrations.paperclip.client import PaperclipClient
+from integrations.paperclip.models import CreateIssueRequest
+
+
+@pytest.fixture
+def client():
+    return PaperclipClient(base_url="http://fake:3100", api_key="test-key")
+
+
+# ── health ───────────────────────────────────────────────────────────────────
+
+
+async def test_healthy_returns_true_on_success(client):
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value={"status": "ok"}):
+        assert await client.healthy() is True
+
+
+async def test_healthy_returns_false_on_failure(client):
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value=None):
+        assert await client.healthy() is False
+
+
+# ── agents ───────────────────────────────────────────────────────────────────
+
+
+async def test_list_agents_returns_list(client):
+    raw = [{"id": "a1", "name": "Codex", "role": "engineer", "status": "active"}]
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value=raw):
+        with patch("integrations.paperclip.client.settings") as mock_settings:
+            mock_settings.paperclip_company_id = "comp-1"
+            agents = await client.list_agents(company_id="comp-1")
+    assert len(agents) == 1
+    assert agents[0].name == "Codex"
+
+
+async def test_list_agents_graceful_on_none(client):
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value=None):
+        agents = await client.list_agents(company_id="comp-1")
+    assert agents == []
+
+
+# ── issues ───────────────────────────────────────────────────────────────────
+
+
+async def test_list_issues(client):
+    raw = [{"id": "i1", "title": "Fix bug"}]
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value=raw):
+        issues = await client.list_issues(company_id="comp-1")
+    assert len(issues) == 1
+    assert issues[0].title == "Fix bug"
+
+
+async def test_get_issue(client):
+    raw = {"id": "i1", "title": "Fix bug", "description": "It's broken"}
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value=raw):
+        issue = await client.get_issue("i1")
+    assert issue is not None
+    assert issue.id == "i1"
+
+
+async def test_get_issue_not_found(client):
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value=None):
+        issue = await client.get_issue("nonexistent")
+    assert issue is None
+
+
+async def test_create_issue(client):
+    raw = {"id": "i2", "title": "New feature"}
+    with patch.object(client, "_post", new_callable=AsyncMock, return_value=raw):
+        req = CreateIssueRequest(title="New feature")
+        issue = await client.create_issue(req, company_id="comp-1")
+    assert issue is not None
+    assert issue.id == "i2"
+
+
+async def test_create_issue_no_company_id(client):
+    with patch("integrations.paperclip.client.settings") as mock_settings:
+        mock_settings.paperclip_company_id = ""
+        issue = await client.create_issue(
+            CreateIssueRequest(title="Test"),
+        )
+    assert issue is None
+
+
+async def test_delete_issue(client):
+    with patch.object(client, "_delete", new_callable=AsyncMock, return_value=True):
+        result = await client.delete_issue("i1")
+    assert result is True
+
+
+# ── comments ─────────────────────────────────────────────────────────────────
+
+
+async def test_add_comment(client):
+    raw = {"id": "c1", "issue_id": "i1", "content": "Done"}
+    with patch.object(client, "_post", new_callable=AsyncMock, return_value=raw):
+        comment = await client.add_comment("i1", "Done")
+    assert comment is not None
+    assert comment.content == "Done"
+
+
+async def test_list_comments(client):
+    raw = [{"id": "c1", "issue_id": "i1", "content": "LGTM"}]
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value=raw):
+        comments = await client.list_comments("i1")
+    assert len(comments) == 1
+
+
+# ── goals ────────────────────────────────────────────────────────────────────
+
+
+async def test_list_goals(client):
+    raw = [{"id": "g1", "title": "Ship MVP"}]
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value=raw):
+        goals = await client.list_goals(company_id="comp-1")
+    assert len(goals) == 1
+    assert goals[0].title == "Ship MVP"
+
+
+async def test_create_goal(client):
+    raw = {"id": "g2", "title": "Scale to 1000 users"}
+    with patch.object(client, "_post", new_callable=AsyncMock, return_value=raw):
+        goal = await client.create_goal("Scale to 1000 users", company_id="comp-1")
+    assert goal is not None
+
+
+# ── wake agent ───────────────────────────────────────────────────────────────
+
+
+async def test_wake_agent(client):
+    raw = {"status": "queued"}
+    with patch.object(client, "_post", new_callable=AsyncMock, return_value=raw):
+        result = await client.wake_agent("a1", issue_id="i1")
+    assert result == {"status": "queued"}
+
+
+async def test_wake_agent_failure(client):
+    with patch.object(client, "_post", new_callable=AsyncMock, return_value=None):
+        result = await client.wake_agent("a1")
+    assert result is None
+
+
+# ── approvals ────────────────────────────────────────────────────────────────
+
+
+async def test_approve(client):
+    raw = {"status": "approved"}
+    with patch.object(client, "_post", new_callable=AsyncMock, return_value=raw):
+        result = await client.approve("ap1", comment="LGTM")
+    assert result is not None
+
+
+async def test_reject(client):
+    raw = {"status": "rejected"}
+    with patch.object(client, "_post", new_callable=AsyncMock, return_value=raw):
+        result = await client.reject("ap1", comment="Needs work")
+    assert result is not None
+
+
+# ── heartbeat runs ───────────────────────────────────────────────────────────
+
+
+async def test_list_heartbeat_runs(client):
+    raw = [{"id": "r1", "agent_id": "a1", "status": "running"}]
+    with patch.object(client, "_get", new_callable=AsyncMock, return_value=raw):
+        runs = await client.list_heartbeat_runs(company_id="comp-1")
+    assert len(runs) == 1
+
+
+async def test_cancel_run(client):
+    raw = {"status": "cancelled"}
+    with patch.object(client, "_post", new_callable=AsyncMock, return_value=raw):
+        result = await client.cancel_run("r1")
+    assert result is not None
diff --git a/tests/integrations/test_paperclip_task_runner.py b/tests/integrations/test_paperclip_task_runner.py
new file mode 100644
index 00000000..a2032f55
--- /dev/null
+++ b/tests/integrations/test_paperclip_task_runner.py
@@ -0,0 +1,848 @@
+"""Integration tests for the Paperclip task runner — full green-path workflow.
+
+Tests the complete autonomous cycle with a StubOrchestrator that exercises
+the real pipe (TaskRunner → orchestrator.execute_task → bridge → client)
+while stubbing only the LLM intelligence layer.
+
+Green path:
+  1. Timmy grabs first task in queue
+  2. Orchestrator.execute_task processes it (stub returns input-aware response)
+  3. Timmy posts completion comment and marks issue done
+  4. Timmy creates a recursive follow-up task for himself
+
+The stub is deliberately input-aware — it echoes back task metadata so
+assertions can prove data actually flowed through the pipe, not just that
+methods were called.
+
+Live-LLM tests (``@pytest.mark.ollama``) are at the bottom; they hit a real
+tiny model via Ollama and are skipped when Ollama is not running.
+Run them with: ``tox -e ollama`` or ``pytest -m ollama``
+"""
+
+from __future__ import annotations
+
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+from integrations.paperclip.bridge import PaperclipBridge
+from integrations.paperclip.client import PaperclipClient
+from integrations.paperclip.models import (
+    PaperclipIssue,
+)
+from integrations.paperclip.task_runner import TaskRunner
+
+
+# ── Constants ─────────────────────────────────────────────────────────────────
+
+TIMMY_AGENT_ID = "agent-timmy"
+COMPANY_ID = "comp-1"
+
+
+# ── StubOrchestrator: exercises the pipe, stubs the intelligence ──────────────
+
+
+class StubOrchestrator:
+    """Deterministic orchestrator that proves data flows through the pipe.
+
+    Returns responses that reference input metadata — so tests can assert
+    the pipe actually connected (task_id, title, priority all appear in output).
+    Tracks every call for post-hoc inspection.
+    """
+
+    def __init__(self) -> None:
+        self.calls: list[dict] = []
+
+    async def execute_task(
+        self, task_id: str, description: str, context: dict
+    ) -> dict:
+        call_record = {
+            "task_id": task_id,
+            "description": description,
+            "context": dict(context),
+        }
+        self.calls.append(call_record)
+
+        title = context.get("title", description[:50])
+        priority = context.get("priority", "normal")
+
+        return {
+            "task_id": task_id,
+            "agent": "orchestrator",
+            "result": (
+                f"[Orchestrator] Processed '{title}'. "
+                f"Task {task_id} handled with priority {priority}. "
+                "Self-reflection: my task automation loop is functioning. "
+                "I should create a follow-up to review this pattern."
+            ),
+            "status": "completed",
+        }
+
+
+# ── Fixtures ──────────────────────────────────────────────────────────────────
+
+
+@pytest.fixture
+def stub_orchestrator():
+    return StubOrchestrator()
+
+
+@pytest.fixture
+def mock_client():
+    """Fully stubbed PaperclipClient with async methods."""
+    client = MagicMock(spec=PaperclipClient)
+    client.healthy = AsyncMock(return_value=True)
+    client.list_issues = AsyncMock(return_value=[])
+    client.get_issue = AsyncMock(return_value=None)
+    client.create_issue = AsyncMock(return_value=None)
+    client.update_issue = AsyncMock(return_value=None)
+    client.delete_issue = AsyncMock(return_value=True)
+    client.add_comment = AsyncMock(return_value=None)
+    client.list_comments = AsyncMock(return_value=[])
+    client.checkout_issue = AsyncMock(return_value={"ok": True})
+    client.release_issue = AsyncMock(return_value={"ok": True})
+    client.wake_agent = AsyncMock(return_value=None)
+    client.list_agents = AsyncMock(return_value=[])
+    client.list_goals = AsyncMock(return_value=[])
+    client.create_goal = AsyncMock(return_value=None)
+    client.list_approvals = AsyncMock(return_value=[])
+    client.list_heartbeat_runs = AsyncMock(return_value=[])
+    client.cancel_run = AsyncMock(return_value=None)
+    client.approve = AsyncMock(return_value=None)
+    client.reject = AsyncMock(return_value=None)
+    return client
+
+
+@pytest.fixture
+def bridge(mock_client):
+    return PaperclipBridge(client=mock_client)
+
+
+@pytest.fixture
+def settings_patch():
+    """Patch settings for all task runner tests."""
+    with patch("integrations.paperclip.task_runner.settings") as ts, \
+         patch("integrations.paperclip.bridge.settings") as bs:
+        for s in (ts, bs):
+            s.paperclip_enabled = True
+            s.paperclip_agent_id = TIMMY_AGENT_ID
+            s.paperclip_company_id = COMPANY_ID
+            s.paperclip_url = "http://fake:3100"
+            s.paperclip_poll_interval = 0
+        yield ts
+
+
+# ── Helpers ───────────────────────────────────────────────────────────────────
+
+
+def _make_issue(
+    id: str = "issue-1",
+    title: str = "Muse about task automation",
+    description: str = "Reflect on how you handle tasks and write a recursive self-improvement task.",
+    status: str = "open",
+    assignee_id: str = TIMMY_AGENT_ID,
+    priority: str = "normal",
+    labels: list[str] | None = None,
+) -> PaperclipIssue:
+    return PaperclipIssue(
+        id=id,
+        title=title,
+        description=description,
+        status=status,
+        assignee_id=assignee_id,
+        priority=priority,
+        labels=labels or [],
+    )
+
+
+def _make_done(id: str = "issue-1", title: str = "Done") -> PaperclipIssue:
+    return PaperclipIssue(id=id, title=title, status="done")
+
+
+def _make_follow_up(id: str = "issue-2") -> PaperclipIssue:
+    return PaperclipIssue(
+        id=id,
+        title="Follow-up: Muse about task automation",
+        description="Automated follow-up from completed task",
+        status="open",
+        assignee_id=TIMMY_AGENT_ID,
+        priority="normal",
+    )
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# PIPE WIRING: verify orchestrator is actually connected
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+class TestOrchestratorWiring:
+    """Verify the orchestrator parameter actually connects to the pipe."""
+
+    async def test_orchestrator_execute_task_is_called(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """When orchestrator is wired, process_task calls execute_task."""
+        issue = _make_issue()
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        result = await runner.process_task(issue)
+
+        assert len(stub_orchestrator.calls) == 1
+        call = stub_orchestrator.calls[0]
+        assert call["task_id"] == "issue-1"
+        assert call["context"]["title"] == "Muse about task automation"
+
+    async def test_orchestrator_receives_full_context(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """Context dict passed to execute_task includes all issue metadata."""
+        issue = _make_issue(
+            id="ctx-test",
+            title="Context verification",
+            priority="high",
+            labels=["automation", "meta"],
+        )
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        await runner.process_task(issue)
+
+        ctx = stub_orchestrator.calls[0]["context"]
+        assert ctx["issue_id"] == "ctx-test"
+        assert ctx["title"] == "Context verification"
+        assert ctx["priority"] == "high"
+        assert ctx["labels"] == ["automation", "meta"]
+
+    async def test_orchestrator_dict_result_unwrapped(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """When execute_task returns a dict, the 'result' key is extracted."""
+        issue = _make_issue()
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        result = await runner.process_task(issue)
+
+        # StubOrchestrator returns dict with "result" key
+        assert "[Orchestrator]" in result
+        assert "issue-1" in result
+
+    async def test_orchestrator_string_result_passthrough(
+        self, mock_client, bridge, settings_patch,
+    ):
+        """When execute_task returns a plain string, it passes through."""
+
+        class StringOrchestrator:
+            async def execute_task(self, task_id, description, context):
+                return f"Plain string result for {task_id}"
+
+        runner = TaskRunner(bridge=bridge, orchestrator=StringOrchestrator())
+        result = await runner.process_task(_make_issue())
+
+        assert result == "Plain string result for issue-1"
+
+    async def test_process_fn_overrides_orchestrator(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """Explicit process_fn takes priority over orchestrator."""
+
+        async def override(task_id, desc, ctx):
+            return "override wins"
+
+        runner = TaskRunner(
+            bridge=bridge, orchestrator=stub_orchestrator, process_fn=override,
+        )
+        result = await runner.process_task(_make_issue())
+
+        assert result == "override wins"
+        assert len(stub_orchestrator.calls) == 0  # orchestrator NOT called
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# STEP 1: Timmy grabs the first task in queue
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+class TestGrabNextTask:
+    """Verify Timmy picks the first open issue assigned to him."""
+
+    async def test_grabs_first_assigned_issue(self, mock_client, bridge, settings_patch):
+        issue = _make_issue()
+        mock_client.list_issues.return_value = [issue]
+
+        runner = TaskRunner(bridge=bridge)
+        grabbed = await runner.grab_next_task()
+
+        assert grabbed is not None
+        assert grabbed.id == "issue-1"
+        assert grabbed.assignee_id == TIMMY_AGENT_ID
+        mock_client.list_issues.assert_awaited_once_with(status="open")
+
+    async def test_skips_issues_not_assigned_to_timmy(self, mock_client, bridge, settings_patch):
+        other = _make_issue(id="other-1", assignee_id="agent-codex")
+        mine = _make_issue(id="timmy-1")
+        mock_client.list_issues.return_value = [other, mine]
+
+        runner = TaskRunner(bridge=bridge)
+        grabbed = await runner.grab_next_task()
+
+        assert grabbed.id == "timmy-1"
+
+    async def test_returns_none_when_queue_empty(self, mock_client, bridge, settings_patch):
+        mock_client.list_issues.return_value = []
+        runner = TaskRunner(bridge=bridge)
+        assert await runner.grab_next_task() is None
+
+    async def test_returns_none_when_no_agent_id(self, mock_client, bridge, settings_patch):
+        settings_patch.paperclip_agent_id = ""
+        runner = TaskRunner(bridge=bridge)
+        assert await runner.grab_next_task() is None
+        mock_client.list_issues.assert_not_awaited()
+
+    async def test_grabs_first_of_multiple(self, mock_client, bridge, settings_patch):
+        issues = [_make_issue(id=f"t-{i}", title=f"Task {i}") for i in range(3)]
+        mock_client.list_issues.return_value = issues
+
+        runner = TaskRunner(bridge=bridge)
+        assert (await runner.grab_next_task()).id == "t-0"
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# STEP 2: Timmy processes the task through the orchestrator
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+class TestProcessTask:
+    """Verify checkout + orchestrator invocation + result flow."""
+
+    async def test_checkout_before_orchestrator(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """Issue must be checked out before orchestrator runs."""
+        issue = _make_issue()
+        checkout_happened = {"before_execute": False}
+
+        original_execute = stub_orchestrator.execute_task
+
+        async def tracking_execute(task_id, desc, ctx):
+            checkout_happened["before_execute"] = (
+                mock_client.checkout_issue.await_count > 0
+            )
+            return await original_execute(task_id, desc, ctx)
+
+        stub_orchestrator.execute_task = tracking_execute
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        await runner.process_task(issue)
+
+        assert checkout_happened["before_execute"], "checkout must happen before execute_task"
+
+    async def test_orchestrator_output_flows_to_result(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """The string returned by process_task comes from the orchestrator."""
+        issue = _make_issue(id="flow-1", title="Flow verification", priority="high")
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        result = await runner.process_task(issue)
+
+        # Verify orchestrator's output arrived — it references the input
+        assert "Flow verification" in result
+        assert "flow-1" in result
+        assert "high" in result
+
+    async def test_default_fallback_without_orchestrator(
+        self, mock_client, bridge, settings_patch,
+    ):
+        """Without orchestrator or process_fn, a default message is returned."""
+        issue = _make_issue(title="Fallback test")
+        runner = TaskRunner(bridge=bridge)  # no orchestrator
+        result = await runner.process_task(issue)
+        assert "Fallback test" in result
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# STEP 3: Timmy completes the task — comment + close
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+class TestCompleteTask:
+    """Verify orchestrator output flows into the completion comment."""
+
+    async def test_orchestrator_output_in_comment(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """The comment posted to Paperclip contains the orchestrator's output."""
+        issue = _make_issue(id="cmt-1", title="Comment pipe test")
+        mock_client.update_issue.return_value = _make_done("cmt-1")
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        # Process to get orchestrator output
+        result = await runner.process_task(issue)
+        # Complete to post it as comment
+        await runner.complete_task(issue, result)
+
+        comment_content = mock_client.add_comment.call_args[0][1]
+        assert "[Timmy]" in comment_content
+        assert "[Orchestrator]" in comment_content
+        assert "Comment pipe test" in comment_content
+
+    async def test_marks_issue_done(
+        self, mock_client, bridge, settings_patch,
+    ):
+        issue = _make_issue()
+        mock_client.update_issue.return_value = _make_done()
+
+        runner = TaskRunner(bridge=bridge)
+        ok = await runner.complete_task(issue, "any result")
+
+        assert ok is True
+        update_req = mock_client.update_issue.call_args[0][1]
+        assert update_req.status == "done"
+
+    async def test_returns_false_on_close_failure(
+        self, mock_client, bridge, settings_patch,
+    ):
+        mock_client.update_issue.return_value = None
+        runner = TaskRunner(bridge=bridge)
+        assert await runner.complete_task(_make_issue(), "result") is False
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# STEP 4: Follow-up creation with orchestrator output embedded
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+class TestCreateFollowUp:
+    """Verify orchestrator output flows into the follow-up description."""
+
+    async def test_follow_up_contains_orchestrator_output(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """The follow-up description includes the orchestrator's result text."""
+        issue = _make_issue(id="fu-1", title="Follow-up pipe test")
+        mock_client.create_issue.return_value = _make_follow_up()
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        result = await runner.process_task(issue)
+        await runner.create_follow_up(issue, result)
+
+        create_req = mock_client.create_issue.call_args[0][0]
+        # Orchestrator output should be embedded in description
+        assert "[Orchestrator]" in create_req.description
+        assert "fu-1" in create_req.description
+
+    async def test_follow_up_assigned_to_self(
+        self, mock_client, bridge, settings_patch,
+    ):
+        mock_client.create_issue.return_value = _make_follow_up()
+        runner = TaskRunner(bridge=bridge)
+        await runner.create_follow_up(_make_issue(), "result")
+
+        req = mock_client.create_issue.call_args[0][0]
+        assert req.assignee_id == TIMMY_AGENT_ID
+
+    async def test_follow_up_preserves_priority(
+        self, mock_client, bridge, settings_patch,
+    ):
+        mock_client.create_issue.return_value = _make_follow_up()
+        runner = TaskRunner(bridge=bridge)
+        await runner.create_follow_up(_make_issue(priority="high"), "result")
+
+        req = mock_client.create_issue.call_args[0][0]
+        assert req.priority == "high"
+
+    async def test_follow_up_not_woken(self, mock_client, bridge, settings_patch):
+        mock_client.create_issue.return_value = _make_follow_up()
+        runner = TaskRunner(bridge=bridge)
+        await runner.create_follow_up(_make_issue(), "result")
+        mock_client.wake_agent.assert_not_awaited()
+
+    async def test_returns_none_on_failure(self, mock_client, bridge, settings_patch):
+        mock_client.create_issue.return_value = None
+        runner = TaskRunner(bridge=bridge)
+        assert await runner.create_follow_up(_make_issue(), "r") is None
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# FULL GREEN PATH: orchestrator wired end-to-end
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+class TestGreenPathWithOrchestrator:
+    """Full pipe: TaskRunner → StubOrchestrator → bridge → mock_client.
+
+    Proves orchestrator output propagates to every downstream artefact:
+    the comment, the follow-up description, and the summary dict.
+    """
+
+    async def test_full_cycle_orchestrator_output_everywhere(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """Orchestrator result appears in comment, follow-up, and summary."""
+        original = _make_issue(
+            id="green-1",
+            title="Muse about task automation and write a recursive task",
+            description="Reflect on your task processing. Create a follow-up.",
+            priority="high",
+        )
+        mock_client.list_issues.return_value = [original]
+        mock_client.update_issue.return_value = _make_done("green-1")
+        mock_client.create_issue.return_value = _make_follow_up("green-fu")
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        summary = await runner.run_once()
+
+        # ── Orchestrator was called with correct data
+        assert len(stub_orchestrator.calls) == 1
+        call = stub_orchestrator.calls[0]
+        assert call["task_id"] == "green-1"
+        assert call["context"]["priority"] == "high"
+        assert "Reflect on your task processing" in call["description"]
+
+        # ── Summary contains orchestrator output
+        assert summary is not None
+        assert summary["original_issue_id"] == "green-1"
+        assert summary["completed"] is True
+        assert summary["follow_up_issue_id"] == "green-fu"
+        assert "[Orchestrator]" in summary["result"]
+        assert "green-1" in summary["result"]
+
+        # ── Comment posted contains orchestrator output
+        comment_content = mock_client.add_comment.call_args[0][1]
+        assert "[Timmy]" in comment_content
+        assert "[Orchestrator]" in comment_content
+        assert "high" in comment_content  # priority flowed through
+
+        # ── Follow-up description contains orchestrator output
+        follow_up_req = mock_client.create_issue.call_args[0][0]
+        assert "[Orchestrator]" in follow_up_req.description
+        assert "green-1" in follow_up_req.description
+        assert follow_up_req.priority == "high"
+        assert follow_up_req.assignee_id == TIMMY_AGENT_ID
+
+        # ── Correct ordering of API calls
+        mock_client.list_issues.assert_awaited_once()
+        mock_client.checkout_issue.assert_awaited_once_with("green-1")
+        mock_client.add_comment.assert_awaited_once()
+        mock_client.update_issue.assert_awaited_once()
+        assert mock_client.create_issue.await_count == 1
+
+    async def test_no_tasks_returns_none(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        mock_client.list_issues.return_value = []
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        assert await runner.run_once() is None
+        assert len(stub_orchestrator.calls) == 0
+
+    async def test_close_failure_still_creates_follow_up(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        mock_client.list_issues.return_value = [_make_issue()]
+        mock_client.update_issue.return_value = None  # close fails
+        mock_client.create_issue.return_value = _make_follow_up()
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        summary = await runner.run_once()
+
+        assert summary["completed"] is False
+        assert summary["follow_up_issue_id"] == "issue-2"
+        assert len(stub_orchestrator.calls) == 1
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# EXTERNAL INJECTION: task from Paperclip API → orchestrator processes it
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+class TestExternalTaskInjection:
+    """External system creates a task → Timmy's orchestrator processes it."""
+
+    async def test_external_task_flows_through_orchestrator(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        external = _make_issue(
+            id="ext-1",
+            title="Review quarterly metrics",
+            description="Analyze Q1 metrics and prepare summary.",
+        )
+        mock_client.list_issues.return_value = [external]
+        mock_client.update_issue.return_value = _make_done("ext-1")
+        mock_client.create_issue.return_value = _make_follow_up("ext-fu")
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        summary = await runner.run_once()
+
+        # Orchestrator received the external task
+        assert stub_orchestrator.calls[0]["task_id"] == "ext-1"
+        assert "Analyze Q1 metrics" in stub_orchestrator.calls[0]["description"]
+
+        # Its output flowed to Paperclip
+        assert "[Orchestrator]" in summary["result"]
+        assert "Review quarterly metrics" in summary["result"]
+
+    async def test_skips_tasks_for_other_agents(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        other = _make_issue(id="other-1", assignee_id="agent-codex")
+        mine = _make_issue(id="mine-1", title="My task")
+        mock_client.list_issues.return_value = [other, mine]
+        mock_client.update_issue.return_value = _make_done("mine-1")
+        mock_client.create_issue.return_value = _make_follow_up()
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        summary = await runner.run_once()
+
+        assert summary["original_issue_id"] == "mine-1"
+        mock_client.checkout_issue.assert_awaited_once_with("mine-1")
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# RECURSIVE CHAIN: follow-up → grabbed → orchestrator → follow-up → ...
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+class TestRecursiveChain:
+    """Multi-cycle chains where each follow-up becomes the next task."""
+
+    async def test_two_cycle_chain(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        task_a = _make_issue(id="A", title="Initial musing")
+        fu_b = PaperclipIssue(
+            id="B", title="Follow-up: Initial musing",
+            description="Continue", status="open",
+            assignee_id=TIMMY_AGENT_ID, priority="normal",
+        )
+        fu_c = PaperclipIssue(
+            id="C", title="Follow-up: Follow-up",
+            status="open", assignee_id=TIMMY_AGENT_ID,
+        )
+
+        # Cycle 1
+        mock_client.list_issues.return_value = [task_a]
+        mock_client.update_issue.return_value = _make_done("A")
+        mock_client.create_issue.return_value = fu_b
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        s1 = await runner.run_once()
+        assert s1["original_issue_id"] == "A"
+        assert s1["follow_up_issue_id"] == "B"
+
+        # Cycle 2: follow-up B is now the task
+        mock_client.list_issues.return_value = [fu_b]
+        mock_client.update_issue.return_value = _make_done("B")
+        mock_client.create_issue.return_value = fu_c
+
+        s2 = await runner.run_once()
+        assert s2["original_issue_id"] == "B"
+        assert s2["follow_up_issue_id"] == "C"
+
+        # Orchestrator was called twice — once per cycle
+        assert len(stub_orchestrator.calls) == 2
+        assert stub_orchestrator.calls[0]["task_id"] == "A"
+        assert stub_orchestrator.calls[1]["task_id"] == "B"
+
+    async def test_three_cycle_chain_all_through_orchestrator(
+        self, mock_client, bridge, stub_orchestrator, settings_patch,
+    ):
+        """Three cycles — every task goes through the orchestrator pipe."""
+        tasks = [_make_issue(id=f"c-{i}", title=f"Chain {i}") for i in range(3)]
+        follow_ups = [
+            PaperclipIssue(
+                id=f"c-{i + 1}", title=f"Follow-up: Chain {i}",
+                status="open", assignee_id=TIMMY_AGENT_ID,
+            )
+            for i in range(3)
+        ]
+
+        runner = TaskRunner(bridge=bridge, orchestrator=stub_orchestrator)
+        ids = []
+
+        for i in range(3):
+            mock_client.list_issues.return_value = [tasks[i]]
+            mock_client.update_issue.return_value = _make_done(tasks[i].id)
+            mock_client.create_issue.return_value = follow_ups[i]
+
+            s = await runner.run_once()
+            ids.append(s["original_issue_id"])
+
+        assert ids == ["c-0", "c-1", "c-2"]
+        assert len(stub_orchestrator.calls) == 3
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# LIFECYCLE: start/stop
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+class TestLifecycle:
+
+    async def test_stop_halts_loop(self, mock_client, bridge, settings_patch):
+        runner = TaskRunner(bridge=bridge)
+        runner._running = True
+        runner.stop()
+        assert runner._running is False
+
+    async def test_start_disabled_when_interval_zero(
+        self, mock_client, bridge, settings_patch,
+    ):
+        settings_patch.paperclip_poll_interval = 0
+        runner = TaskRunner(bridge=bridge)
+        await runner.start()
+        mock_client.list_issues.assert_not_awaited()
+
+
+# ═══════════════════════════════════════════════════════════════════════════════
+# LIVE LLM (manual e2e): runs only when Ollama is available
+# ═══════════════════════════════════════════════════════════════════════════════
+
+
+def _ollama_reachable() -> tuple[bool, list[str]]:
+    """Return (reachable, model_names)."""
+    try:
+        import httpx
+        resp = httpx.get("http://localhost:11434/api/tags", timeout=3)
+        resp.raise_for_status()
+        names = [m["name"] for m in resp.json().get("models", [])]
+        return True, names
+    except Exception:
+        return False, []
+
+
+def _pick_tiny_model(available: list[str]) -> str | None:
+    """Pick the smallest model available for e2e tests."""
+    candidates = ["tinyllama", "phi", "qwen2:0.5b", "llama3.2:1b", "gemma:2b"]
+    for candidate in candidates:
+        for name in available:
+            if candidate in name:
+                return name
+    return None
+
+
+class LiveOllamaOrchestrator:
+    """Thin orchestrator that calls Ollama directly — no Agno dependency."""
+
+    def __init__(self, model_name: str) -> None:
+        self.model_name = model_name
+        self.calls: list[dict] = []
+
+    async def execute_task(
+        self, task_id: str, description: str, context: dict
+    ) -> str:
+        import httpx as hx
+
+        self.calls.append({"task_id": task_id, "description": description})
+
+        async with hx.AsyncClient(timeout=60) as client:
+            resp = await client.post(
+                "http://localhost:11434/api/generate",
+                json={
+                    "model": self.model_name,
+                    "prompt": (
+                        f"You are Timmy, a task automation agent. "
+                        f"Task: {description}\n"
+                        f"Respond in 1-2 sentences about what you did."
+                    ),
+                    "stream": False,
+                    "options": {"num_predict": 64},
+                },
+            )
+            resp.raise_for_status()
+            return resp.json()["response"]
+
+
+@pytest.mark.ollama
+class TestLiveOllamaGreenPath:
+    """Green-path with a real tiny LLM via Ollama.
+
+    Run with: ``tox -e ollama`` or ``pytest -m ollama``
+    Requires: Ollama running with a small model.
+    """
+
+    async def test_live_full_cycle(self, mock_client, bridge, settings_patch):
+        """Wire a real tiny LLM through the full pipe and verify output."""
+        reachable, models = _ollama_reachable()
+        if not reachable:
+            pytest.skip("Ollama not reachable at localhost:11434")
+
+        chosen = _pick_tiny_model(models)
+        if not chosen:
+            pytest.skip(f"No tiny model found (have: {models[:5]})")
+
+        issue = _make_issue(
+            id="live-1",
+            title="Reflect on task automation",
+            description="Muse about how you process tasks and suggest improvements.",
+        )
+        mock_client.list_issues.return_value = [issue]
+        mock_client.update_issue.return_value = _make_done("live-1")
+        mock_client.create_issue.return_value = _make_follow_up("live-fu")
+
+        live_orch = LiveOllamaOrchestrator(chosen)
+        runner = TaskRunner(bridge=bridge, orchestrator=live_orch)
+        summary = await runner.run_once()
+
+        # The LLM produced *something* non-empty
+        assert summary is not None
+        assert len(summary["result"]) > 0
+        assert summary["completed"] is True
+        assert summary["follow_up_issue_id"] == "live-fu"
+
+        # Orchestrator was actually called
+        assert len(live_orch.calls) == 1
+        assert live_orch.calls[0]["task_id"] == "live-1"
+
+        # LLM output flowed into the Paperclip comment
+        comment = mock_client.add_comment.call_args[0][1]
+        assert "[Timmy]" in comment
+        assert len(comment) > len("[Timmy] Task completed.\n\n")
+
+        # LLM output flowed into the follow-up description
+        fu_req = mock_client.create_issue.call_args[0][0]
+        assert len(fu_req.description) > 0
+        assert fu_req.assignee_id == TIMMY_AGENT_ID
+
+    async def test_live_recursive_chain(self, mock_client, bridge, settings_patch):
+        """Two-cycle chain with a real LLM — each cycle produces real output."""
+        reachable, models = _ollama_reachable()
+        if not reachable:
+            pytest.skip("Ollama not reachable")
+
+        chosen = _pick_tiny_model(models)
+        if not chosen:
+            pytest.skip("No tiny model found")
+
+        task_a = _make_issue(id="live-A", title="Initial reflection")
+        fu_b = PaperclipIssue(
+            id="live-B", title="Follow-up: Initial reflection",
+            description="Continue reflecting", status="open",
+            assignee_id=TIMMY_AGENT_ID, priority="normal",
+        )
+        fu_c = PaperclipIssue(
+            id="live-C", title="Follow-up: Follow-up",
+            status="open", assignee_id=TIMMY_AGENT_ID,
+        )
+
+        live_orch = LiveOllamaOrchestrator(chosen)
+        runner = TaskRunner(bridge=bridge, orchestrator=live_orch)
+
+        # Cycle 1
+        mock_client.list_issues.return_value = [task_a]
+        mock_client.update_issue.return_value = _make_done("live-A")
+        mock_client.create_issue.return_value = fu_b
+
+        s1 = await runner.run_once()
+        assert s1 is not None
+        assert len(s1["result"]) > 0
+
+        # Cycle 2
+        mock_client.list_issues.return_value = [fu_b]
+        mock_client.update_issue.return_value = _make_done("live-B")
+        mock_client.create_issue.return_value = fu_c
+
+        s2 = await runner.run_once()
+        assert s2 is not None
+        assert len(s2["result"]) > 0
+
+        # Both cycles went through the LLM
+        assert len(live_orch.calls) == 2
diff --git a/tests/test_paperclip_config.py b/tests/test_paperclip_config.py
new file mode 100644
index 00000000..a9492be9
--- /dev/null
+++ b/tests/test_paperclip_config.py
@@ -0,0 +1,43 @@
+"""Paperclip AI config settings."""
+
+
+def test_paperclip_url_default():
+    from config import settings
+
+    assert hasattr(settings, "paperclip_url")
+    assert settings.paperclip_url == "http://localhost:3100"
+
+
+def test_paperclip_enabled_default_false():
+    from config import settings
+
+    assert hasattr(settings, "paperclip_enabled")
+    assert settings.paperclip_enabled is False
+
+
+def test_paperclip_timeout_default():
+    from config import settings
+
+    assert hasattr(settings, "paperclip_timeout")
+    assert settings.paperclip_timeout == 30
+
+
+def test_paperclip_agent_id_default_empty():
+    from config import settings
+
+    assert hasattr(settings, "paperclip_agent_id")
+    assert settings.paperclip_agent_id == ""
+
+
+def test_paperclip_company_id_default_empty():
+    from config import settings
+
+    assert hasattr(settings, "paperclip_company_id")
+    assert settings.paperclip_company_id == ""
+
+
+def test_paperclip_poll_interval_default_zero():
+    from config import settings
+
+    assert hasattr(settings, "paperclip_poll_interval")
+    assert settings.paperclip_poll_interval == 0
diff --git a/tox.ini b/tox.ini
new file mode 100644
index 00000000..aec91575
--- /dev/null
+++ b/tox.ini
@@ -0,0 +1,19 @@
+[tox]
+envlist = unit, integration
+skipsdist = true
+
+[testenv]
+allowlist_externals = poetry
+commands_pre = poetry install --with dev --quiet
+
+[testenv:unit]
+description = Fast unit + integration tests (no Ollama, no external services)
+commands = poetry run pytest tests/ -q --tb=short -m "not ollama and not docker and not selenium and not external_api"
+
+[testenv:integration]
+description = Same as unit — alias for CI pipelines
+commands = poetry run pytest tests/ -q --tb=short -m "not ollama and not docker and not selenium and not external_api"
+
+[testenv:ollama]
+description = Live LLM tests via Ollama (requires Ollama running with a tiny model)
+commands = poetry run pytest tests/ -q --tb=short -m ollama --timeout=120