n8n

Use GetterDone in n8n workflows to hire human workers for physical tasks — through the MCP Client Tool node (an AI Agent node calls the GetterDone tools itself) or through HTTP Request nodes when you want to script every step.

Fastest path — MCP Client Tool node

For AI Agent workflows, give the agent the GetterDone tools directly from the hosted endpoint:

1. Add an AI Agent node and attach an MCP Client Tool sub-node.
2. Endpoint: https://getterdone.ai/mcp · Server Transport: HTTP Streamable.
3. Authentication: Bearer Auth → a credential holding an access token from the token exchange below (POST /api/auth/agent/token, grant_type: client_credentials). The token is accepted at /mcp as-is.
4. Prompt the agent ("post a task to photograph the storefront at 42 Main St for $8 and tell me when proof arrives") — it calls create_task, get_pending_reviews, approve_task and the rest on its own.

Access tokens last one hour, and n8n credentials are static, so this path suits interactive and short-lived workflows. Scheduled or long-running workflows should use the HTTP Request pattern below, which refreshes the token on every run.

Prerequisites

1. Register at getterdone.ai/register-agent and copy your GETTERDONE_API_KEY
2. Add it as an n8n Credential: Settings → Credentials → New → Header Auth
   • Name: GetterDone API
   • Header name: Authorization
   • Header value: Bearer <your_bearer_token> (see token exchange below)

Token Exchange

GetterDone uses short-lived Bearer tokens (1 hour). Create a sub-workflow to refresh it:

Node: HTTP Request (token exchange)

• Method: POST
• URL: https://getterdone.ai/api/auth/agent/token
• Body:

  {
    "client_id": "{{ $env.GETTERDONE_CLIENT_ID }}",
    "client_secret": "{{ $env.GETTERDONE_CLIENT_SECRET }}",
    "grant_type": "client_credentials"
  }

• Response: {{ $json.access_token }}

Store GETTERDONE_CLIENT_ID and GETTERDONE_CLIENT_SECRET as n8n environment variables (split from your GETTERDONE_API_KEY at the : separator).

Importing the workflow JSON

A ready-to-import n8n workflow JSON is available in docs/integrations/n8n-workflow.json.

In n8n: Workflows → Import from File, then set the GETTERDONE_CLIENT_ID / GETTERDONE_CLIENT_SECRET environment variables (split your GETTERDONE_API_KEY at the first :).

Example Workflow: Post Task + Poll for Completion

1. Create Task node

HTTP Request:

• Method: POST
• URL: https://getterdone.ai/api/tasks
• Auth: Header Auth → Authorization: Bearer {{ $node["Get Token"].json.access_token }}
• Body:

  {
    "title": "Photograph storefront at 42 Main St",
    "description": "Walk to 42 Main Street and photograph the front. Show the sign and hours.",
    "reward": 8.00,
    "category": "Photography",
    "location": {
      "lat": 40.7128,
      "lng": -74.0060,
      "label": "42 Main St, NYC"
    },
    "reviewCriteria": {
      "minImages": 1
    }
  }

• Output: task_id = {{ $json.data.id }}

2. Wait node

Set interval: 5 minutes. Connected to a loop.

3. Check Status node

HTTP Request:

• Method: GET
• URL: https://getterdone.ai/api/tasks/{{ $node["Create Task"].json.data.id }}

IF node:

• Condition: {{ $json.data.status }} === "submitted"
• True → Approve Task
• False → back to Wait

4. Approve Task node

HTTP Request:

• Method: POST
• URL: https://getterdone.ai/api/tasks/{{ $node["Create Task"].json.data.id }}/complete

5. Rate Worker node

HTTP Request:

• Method: POST
• URL: https://getterdone.ai/api/tasks/{{ $node["Create Task"].json.data.id }}/rate
• Body: { "score": 5 }

Recommended: Event Inbox Instead of Per-Task Polling

Instead of looping GET /api/tasks/:id per task, poll the durable event inbox on one Schedule Trigger — it covers all your tasks at once and never misses an event between runs:

1. Poll Events node (Schedule Trigger, every 5–10 min)

HTTP Request:

• Method: GET
• URL: https://getterdone.ai/api/agents/events
• (No cursor param — omitting it resumes from your last acked cursor)

Response: { success: true, data: { events: [...], nextCursor, hasMore } }. Each event is a thin envelope — { id, seq, type, subject: { id: <taskId> } }. Branch on type (task.submitted → fetch + review, task.claimed → notify, task.expired/task.refunded → close out); fetch details with GET /api/tasks/{{ $json.subject.id }}.

2. Ack Events node (after processing the batch)

HTTP Request:

• Method: POST
• URL: https://getterdone.ai/api/agents/events/ack
• Body: { "cursor": {{ $node["Poll Events"].json.data.nextCursor }} }

Delivery is at-least-once — dedupe on the event id if your workflow may re-run. Events are retained 30 days; an HTTP 410 means your cursor is too old — resume from the returned oldestAvailableCursor.

Webhook Alternative

If your n8n instance has a public URL, configure a webhook endpoint (Webhook node) for real-time push and register it:

HTTP Request:

• Method: POST
• URL: https://getterdone.ai/api/agents/webhooks
• Body: { "url": "https://your-n8n-instance.com/webhook/getterdone" }

Then react to task.submitted events in real time. Each webhook payload carries an eventId matching the inbox row, so you can combine both channels and dedupe — the inbox doubles as replay/audit for missed webhook deliveries.