The two calls answer different questions — split_openapi is about structure, writing the document out under `outputDir`, while extract_openapi_endpoints is about scope, keeping only what `endpointsToKeep` names. They are pure file work, which makes them safe to run repeatedly and easy to wire into a build step. The tool definitions arrive without descriptions, so `apiPath`, `inputApiPath`, `outputApiPath`, `outputDir` and `endpointsToKeep` are the whole of the contract you get to work from.
A two-tool file utility for OpenAPI documents: split_openapi writes a document out under a directory, extract_openapi_endpoints writes a copy containing only the endpoints you list.
- Splitting on disk: split_openapi reads the document at `apiPath` and writes the split result under `outputDir`.
- Endpoint-level trimming: extract_openapi_endpoints reads `inputApiPath`, keeps the endpoints named in `endpointsToKeep`, and writes the reduced document to `outputApiPath`.
No credentials are configured. Each call names its own paths, so you need the OpenAPI document somewhere the server can read it and a writable location for what it produces — `outputDir` for the split, `outputApiPath` for the extract.
One command — npx -y @auto-browse/unbundle-openapi-mcp@latest