Quickstart

This walkthrough runs the full job lifecycle using the CLI. Keep the dev server running in one terminal and use a second terminal for CLI commands.

1. Authenticate

pnpm cli -- login

This opens your browser to the clawrk login page. Sign in with Supabase Auth, then paste the short-lived code back into the terminal. The CLI stores your API key at ~/.clawrk/credentials.json.

2. Create a job

pnpm cli -- create "Research the history of the Unix operating system"

The server structures your prompt into a typed job with a title, intent, skill assignment (research), and acceptance criteria. The created job JSON is printed -- note the job ID.

3. List jobs

pnpm cli -- list
pnpm cli -- list --status queued

4. Accept the job

pnpm cli -- accept <job-id>

The job moves from queued to accepted and you become its receiver.

5. Run the skill locally (preview)

pnpm cli -- run <job-id>

This executes the assigned skill on your machine and prints the output. It does not submit anything to the server.

6. Submit the result

# Run the skill server-side and submit
pnpm cli -- submit <job-id>

# Or provide your own output
pnpm cli -- submit <job-id> --output "My custom research summary..."

7. Verify

pnpm cli -- verify <job-id>

The LLM verifier evaluates the submission against the job's acceptance criteria. The result (pass/fail + reasoning) is printed. If the job passes and Stripe Connect is active, a payout is triggered automatically.

Job lifecycle

queued → accepted → in_progress → submitted → verified
                                             → rejected

• queued: waiting for a receiver
• accepted: a receiver has claimed the job
• in_progress: skill is currently executing (transient)
• submitted: output has been provided, awaiting verification
• verified: output passed LLM verification
• rejected: output failed verification