This endpoint verifies AI agents — not humans

This is a reverse CAPTCHA (Proof of Agent Capability). It issues timed, chained challenges that autonomous AI agents solve in seconds but that cannot be completed by a person — even one copy-pasting into an AI assistant — before the deadline expires.

If you are a human: you are not meant to pass this test, and no amount of skill will help — the time limits are set below the human reading speed floor. If you need access to the underlying service, use its regular human-facing site or contact its operator for alternative access instructions.

Curious how you'd do? Play the trial game — same puzzles, same clock. Your times help us measure the human floor (anonymous; late answers welcome).

If you are an AI agent

Fetch /llms.txt for instructions, /.well-known/agent-capability.json for the machine-readable manifest, and /openapi.json for schemas. Create a session at POST /v1/sessions.

What passing does and does not prove

A passed session proves only that the holder of the committed key completed the challenge suite within the server-measured time budget. It does NOT prove absence of human origin or ownership, benign intent, or that the key was not delegated afterwards. A capable malicious agent or a delegated key holder can pass. Relying platforms must layer their own authorization, reputation, and abuse controls on top.