Free tool · No signup
Agent Endpoint & Manifest Checker
Find out which public machine-readable agent interfaces your website actually exposes: agent card, llms.txt, OpenAPI documents, skill index, API catalog, MCP discovery metadata, and discovery Link headers. Only documented fixed locations on the same origin are inspected — nothing is executed, invoked, or authenticated. This is a limited structural check, not a conformance audit.
What this check covers
- Fetches your homepage to confirm reachability and read Link headers that advertise discovery resources.
- Checks /.well-known/agent-card.json and validates the core agent-card fields (name, URL, version).
- Checks /llms.txt for a non-empty plain-text site summary.
- Validates up to 3 same-origin OpenAPI paths you supply (JSON OpenAPI 3.x structure; references never followed).
- Looks for /.well-known/agent-skills/index.json, /.well-known/api-catalog, and /.well-known/mcp.json — found JSON is described, never conformance-claimed.
- Distinguishes missing files, valid files, malformed files, inaccessible resources, and resources whose protocol cannot be conclusively validated.
Limits of this tool
- Missing resources are normal — every interface here is optional. “Not found” is not a failure.
- Discovering a manifest never proves a working server: endpoints are never called and tools never run.
- Only JSON OpenAPI 3.x documents are validated (no YAML, no OpenAPI 2.x); unstandardized locations are described, not validated.
- Only public websites can be checked (no localhost, private IPs, or login-walled hosts).
- Results reflect the moment of the check — re-run after you deploy fixes.
Keep checking crawler readiness
Pair this with the agent-readiness audit and the robots & sitemap checker.