Agent Readiness Check
AI agents do not read a site the way people do. They look for machine entry points: llms.txt, an OpenAPI description, an MCP server, a feed of changes, a way to pay per call, and a homepage that does not block them. This check runs the 15 tests New Runtime uses on itself and on the 30 services we benchmarked.
Public GET requests plus one MCP handshake; no login, nothing stored except a 24-hour cached result. Agents can call GET /agent-check.json?url=example.com (202 while running, then 200).
| Check | Stage | Result | Evidence |
|---|
Fixes, highest gain first
Badge
Show the score on your README or site; it refreshes from the cached result.
Limits: the checks mirror the surfaces New Runtime built, so they favour its choices; documentation, SDKs or an OpenAPI file hosted elsewhere are not seen. Want the full audit with a 7-day plan and a before/after measurement? Contact us.