MCPHub

What is MCPHub?

MCPHub is an embeddable Model Context Protocol (MCP) solution for AI services that enables seamless integration of MCP servers into AI frameworks like OpenAI Agents, LangChain, and Autogen. Developers can configure, set up, and manage MCP servers within their applications using a unified JSON-based configuration.

How to use MCPHub?

Install with pip install mcphub (optionally with framework-specific extras like [openai]). Create a .mcphub.json file in your project root specifying the MCP servers. Then use the Python API – initialize MCPHub(), call methods like fetch_openai_mcp_server() with the server name, and pass the returned server object to an AI framework (e.g., OpenAI Agents’ Agent with mcp_servers=[server]).

Key features of MCPHub

• JSON-based configuration in .mcphub.json
• Supports TypeScript (npx) and Python (uv) MCP servers
• Automatic server setup from GitHub repositories
• Framework adapters for OpenAI Agents, LangChain, Autogen
• stdio and SSE transport support
• Tool discovery, listing, and optional caching for performance

Use cases of MCPHub

• Add MCP tools to an OpenAI Agent for complex reasoning tasks
• Use MCP tools as LangChain tools in a chain or agent
• Convert MCP servers into Autogen adapters for multi-agent systems
• Manage multiple MCP servers from a single configuration file
• Run MCP servers with SSE endpoints for web-based real-time communication

FAQ from MCPHub

What are the prerequisites for using MCPHub?

You need Python (with uv package manager), git, and npx (comes with Node.js). Install them as shown in the README.

How do I add a new MCP server?

You can either add the server configuration directly to .mcphub.json or use the add_server_from_repo method to automatically configure a server from its GitHub repository (requires an OpenAI API key set as OPENAI_API_KEY).

What transport protocols does MCPHub support?

MCPHub supports both stdio transport (servers run as local subprocesses) and SSE transport (servers run with Server-Sent Events, enabled via mcphub run --sse). SSE provides endpoints at /sse and /message.

Can I use environment variables in the server configuration?

Yes. In .mcphub.json, you can set environment variables per server using the env field, and reference existing environment variables with ${VARIABLE_NAME} syntax.

What AI frameworks are supported?

MCPHub provides adapters for OpenAI Agents, LangChain, and Autogen. Framework-specific dependencies can be installed with extras like mcphub[openai], mcphub[langchain], or mcphub[autogen].