MCP Client

Example Use Cases

• 1

  Connect to MCP servers and enumerate or invoke server-provided tools over stdio or SSE.

• 2

  Perform OpenAI-style function calling to let models automatically select and execute external tools with configurable model parameters.

• 3

  Integrate with custom OpenAI-compatible endpoints (e.g., Anthropic, Azure, Alibaba) while managing retries, timeouts, and server configuration via JSON.

Description

This repository provides a general-purpose MCP (Model Context Protocol) client for Python that can connect to MCP servers over stdio or SSE and expose server-provided tools to language models. It integrates with OpenAI-compatible APIs (and other providers) to perform function-calling and can automatically select and execute tools based on LLM responses. The client is configurable via JSON files, supports custom endpoints and model parameters, and includes an example weather MCP server for testing. It is distributed under the Apache 2.0 license and intended as an open-source SDK for building agent/tooling workflows.