MCP Server: Scalable OpenAPI Endpoint Discovery and API Request Tool

What is MCP Server: Scalable OpenAPI Endpoint Discovery and API Request Tool?

This MCP server enables natural-language discovery and invocation of REST API endpoints from a remote OpenAPI JSON specification. It uses in‑memory semantic search to find relevant endpoints even in large specs (100 KB+), then constructs and executes the corresponding HTTP requests. Designed for developers who want Claude Desktop or other MCP clients to interact with private, large OpenAPI‑defined APIs.

How to use MCP Server: Scalable OpenAPI Endpoint Discovery and API Request Tool?

Install via pip install mcp-server-any-openapi or run the Docker image buryhuang/mcp-server-any-openapi:latest. Configure with environment variables: OPENAPI_JSON_DOCS_URL (remote OpenAPI JSON), MCP_API_PREFIX (tool naming), and optionally GLOBAL_TOOL_PROMPT to help the model select the right tools. Then register the server in your MCP client (e.g., Claude Desktop) using the Docker command template provided in the README. The server exposes two tools: {prefix}_api_request_schema to discover endpoints and {prefix}_make_request to execute them.

Key features of MCP Server: Scalable OpenAPI Endpoint Discovery and API Request Tool

• 🧠 Remote OpenAPI JSON as source – no local file updates needed
• 🔍 Semantic search with optimized MiniLM‑L3 model (43 MB)
• ⚡ In‑memory FAISS vector search for instant endpoint discovery
• 🧠 Endpoint‑based chunking handles 100 KB+ OpenAPI specs losslessly
• 🚀 Async FastAPI server with MCP protocol support
• 🐳 Multi‑architecture Docker images (linux/amd64, linux/arm64)

Use cases of MCP Server: Scalable OpenAPI Endpoint Discovery and API Request Tool

• Ask Claude to “get user profile” or “create a new job posting” against a large private API
• Serve multiple independent API sets by running separate server instances with different prefixes
• Automate API testing by letting an MCP client discover and invoke endpoints from natural‑language instructions
• Override base URLs for staging or dev environments without changing the OpenAPI spec

FAQ from MCP Server: Scalable OpenAPI Endpoint Discovery and API Request Tool

How does this server differ from standard “fetch” or “make request” tools?

Standard tools often fail when the OpenAPI spec is hundreds of kilobytes. This server uses endpoint‑centric semantic search to find the correct endpoint and returns full parameter context, then makes the actual request. It was created specifically to solve the “spec too large” problem.

What are the runtime dependencies and cold start time?

Requires Python or Docker. The first start may need to download the MiniLM‑L3 model (pre‑embedded in the Docker image). Cold start penalty is roughly 15 seconds for model loading if not using the Docker image.

Can I run multiple API sets simultaneously?

Yes. Run multiple instances of the server, each with a different MCP_API_PREFIX and OPENAPI_JSON_DOCS_URL. Example configurations for “finance” and “healthcare” APIs are provided in