PipedAI/ docsSign in

Quickstart

Sign up, register a worker, and fire your first run in about five minutes.

PipedAI has three pieces: a dashboard you use, an API that orchestrates, and a worker process you run on a machine you control. This guide walks through the end-to-end happy path.

1. Create an account

Sign up at /signup with email + password. You'll be dropped into a one-screen wizard at /setup that creates your first workspace and environment in one shot — name them, slug them, submit.

2. Install the worker

On any machine you control (laptop, server, EC2 — anywhere with Node 22+ and a Claude Code install):

bashnpm i -g @pipedai/worker

Verify the install with the doctor command. It runs seven checks (claude on PATH, auth, network, file permissions) and prints pass/fail with hints.

bashpipedai-worker doctor

See the worker install guide for OS-specific service templates (systemd, launchd, nssm, Docker).

3. Generate a registration token

In the dashboard, navigate to Settings → API Keys → Worker Registration (wrk_). Click Create wrk_ key, give it a name (the machine's hostname is fine), and copy the token from the one-time dialog — you can't retrieve it after.

4. Register the worker

bashpipedai-worker register \
  --token wrk_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx \
  --api-url https://api-beta.pipedai.app

Output: Worker registered. workerId: …. The worker writes ~/.pipedai/worker.json with its long-lived wkt_ token (chmod 600). The original wrk_ token is consumed and can be revoked from the dashboard.

Auth modes
Default is claude-max — the worker uses the Claude Max account that claude login sat on the machine. For api-key mode, set AUTH_MODE=api-key and ANTHROPIC_API_KEY=sk-… before registering and running.

5. Author your first trigger

In the dashboard, go to Triggers → New Trigger. Fill out:

  1. Name: anything memorable
  2. Worker: the worker you just registered (sticky assignment)
  3. Prompt: what you want Claude to do
  4. Schedule: cron expression (e.g. 0 9 * * 1-5) or on-demand only
  5. MCP URL: your MCP endpoint
  6. MCP headers: a JSON object with the Authorization header for the MCP service token. Encrypted at rest.

6. Start the worker

bashpipedai-worker run

The worker polls every 60 seconds (configurable via POLL_INTERVAL_MS) and executes any due runs. For long-running production use, install it as a system service — see the worker install guide for systemd / launchd / nssm / Docker templates.

7. Watch the run history

On the trigger detail page, the “Run history” surface shows every firing with status, duration, token usage, and expandable stdout/stderr. The workspace overview at /w/<wsId>/e/<envId> rolls up online workers, today's runs, success rate, and token spend.

Next steps

  1. MCP authoring guide — service tokens, common patterns
  2. Completion webhooks — get notified when a run finishes (HMAC-signed)
  3. Bring your own machine — when to use it, how cost works
  4. Troubleshooting — when something goes wrong