Summary
Big specs become navigable rather than unreadable.
A serious OpenAPI document is far too large to hand a model in one piece, and truncating it loses exactly the endpoint you needed. Walking it in layers keeps the whole API reachable at a fraction of the context.
What it is
A navigator for one OpenAPI specification. It loads the spec from a URL or a file and lets a model walk it in layers — categories, then endpoints, then the detail of a single operation or schema.
What you get
- Top-level information about the loaded specification
- The categories the API is organised into
- The endpoints listed rather than dumped
- One endpoint in full — parameters, request and response
- One schema in full
- Remote specs cached locally, so repeat exploration does not refetch
Requirements
Docker, or a downloaded binary for Linux amd64 or arm64. OPENAPI_SPEC_URL is required and takes a URL or a file path; the cache location is optional and defaults to ~/.openapi-mcp-cache. Stdio, HTTP and interactive CLI modes ship as separate builds.
Setup effort
One command — docker pull ghcr.io/sagenkoder/go-openapi-exploration-mcp-server:latest