FastAPI
Exposes a single tool via the OpenAI Model Context Protocol (MCP) using JSON-RPC 2.0 over HTTP.
About
This FastAPI-based server provides a minimal yet production-ready implementation of an MCP server. It's designed for deployment on Google Cloud Run and demonstrates the OpenAI Model Context Protocol by exposing a `get_time` tool. The server adheres strictly to the MCP specification for `tools/list` and `tools/call`, handling both human-readable content and machine-parseable structuredContent, complete with proper JSON-RPC error codes and input validation. It includes features like CORS support, optional API key authentication, a health check endpoint, and simple request logging, making it a robust foundation for integrating custom tools with AI agents.
Key Features
- Implements `tools/list` and `tools/call` to MCP specification
- Returns both content (text) and structuredContent (JSON)
- Includes JSON-RPC error codes and jsonschema input validation
- CORS enabled and optional API key authentication (x-api-key)
- Dockerfile and one-command Google Cloud Run deploy
- 0 GitHub stars
Use Cases
- Integrating a custom tool into an OpenAI Agent Builder
- Developing and testing custom tools compatible with the OpenAI Model Context Protocol
- Deploying a lightweight, serverless tool endpoint on Google Cloud Run
