This repository provides a scalable solution for OpenAPI endpoint discovery and API requests, utilizing in-memory semantic search and a FastAPI server to efficiently serve API documentation to MCP clients.
This project addresses the challenge of using large OpenAPI specifications with tools like Claude. It uses in-memory semantic search to find relevant API endpoints based on natural language queries, returning complete endpoint documentation in milliseconds. This allows Claude to understand which API to call with full parameters. It also includes a tool to make the actual restful request.
Key Features: Utilizes remote OpenAPI JSON files, semantic search with an optimized MiniLM-L3 model, and a FastAPI-based server with async support. It employs endpoint-based chunking for large OpenAPI specs and in-memory FAISS vector search for fast endpoint discovery.
Usage: Designed for integrating with Claude Desktop, it provides tools for retrieving API schemas and making requests, ensuring reliable execution with complex APIs. The server can be configured with multiple instances for different sets of APIs.
baryhuang/mcp-server-any-openapi
February 9, 2025
March 28, 2025
Python