Free agency tools / MCP Server Health Checker
Does your MCP server actually work?
A 200 response doesn't mean an AI client can use your MCP server. We run the real handshake — initialize then tools/list— and tell you what it advertises, or exactly where it fails.
Endpoint
Advertised tools
Handshake steps
How it works
- 1Paste your MCP endpoint URL (Streamable HTTP), e.g. https://example.com/mcp.
- 2We send a real initialize handshake (protocol 2025-06-18) and capture the server info, capabilities and session id.
- 3Then tools/list runs to see what the server advertises, and you get a verdict with per-step latency.
Frequently asked
What do the verdicts mean?
GOOD = speaks MCP and lists tools. WARN = speaks MCP but exposes no tools. AUTH = the server needs authentication (401/403). DOWN = unreachable or didn't return a valid MCP response.
Does it support SSE servers?
It targets Streamable HTTP (the current transport) and also reads Server-Sent-Events responses. Older SSE-only servers that require a separate GET /sse stream may report DOWN — point it at the Streamable HTTP endpoint.
Do you store my endpoint or tokens?
No credentials are sent, and the check is a public read of the handshake. Authenticated servers correctly report AUTH.