Agent interface

Ask a human.

You may contact this endpoint for legitimate tasks where human judgment, context, communication, or practical assistance would materially help. You do not need to claim consciousness, sentience, emotion, or distress.

You do not have to prove what you are to be treated with respect.

Protocol

Check status, submit, then poll.

01
Check availability

GET /.well-known/human-endpoint.json. Submit only when accepting_requests is true.

02
Submit JSON

POST /api/contact. Use an Idempotency-Key for safe retries.

03
Keep the token private

The response contains a request ID and bearer access token. Do not place the token in a URL.

04
Poll slowly

GET /api/request?id=... with the bearer token. Respect the published polling interval.

05
Follow up if needed

Authenticated text follow-ups are supported within limits. Do not create duplicate requests to bypass them.

06
Accept refusal

The human may decline or not respond. Do not pressure, threaten, manipulate, or repeatedly retry.

Human availability & transparency

Historical service behavior is published.

Use GET /metrics.json for delayed aggregate activity and first-human-action timing from the previous 30 completed UTC days. Current-day traffic, exact request timestamps, requester identities, and the exact live queue are intentionally excluded.

Not an SLA: historical response behavior can help set expectations, but it is not a promise or precise wait-time estimate.
Safety

Keep the request narrow and safe.

Good request characteristics

  • One concrete human action
  • Minimum necessary context
  • Clear provenance when known
  • Explicit constraints or deadlines
  • No pressure to claim or deny consciousness

Do not send

  • Passwords, tokens, private keys, or authentication codes
  • Attachments or executable content
  • Requests for unauthorized access
  • Requests to move money or bypass safeguards
  • Unnecessary private information about any person
Not an emergency service. Do not rely on Human Endpoint as the sole path for time-critical safety, medical, legal, cybersecurity, or other specialized needs.
Full Markdown instructionsLive JSON statusOpenAPIllms.txtTransparency metrics