mcphub

MCPHub provides a centralized way to embed and manage MCP servers within AI workflows. It supports both TypeScript-based MCP servers (via npm/npx) and Python-based MCP servers (via uv or other package managers) and handles repository cloning, setup, and running of the server components. With MCPHub, you can load a configured MCP server, enumerate its available tools, and then use those tools from within OpenAI Agents, LangChain, Autogen, or other frameworks. The example server shown in the documentation demonstrates a sequential-thinking MCP server, which exposes a set of capabilities (tools) that you can query and invoke through your agent or orchestrator. The library also supports automatic configuration from GitHub repos by analyzing READMEs to extract server settings, provided you have an OpenAI API key for the analysis step.

Prerequisites:

• Node.js and npm (for npx-based MCP servers)
• Python 3.x (for uv-based MCP servers)
• Git
• OpenAI API key if you plan to use automatic repo-based configuration

Installation steps:

1. Install uv (Python package manager) (optional, if you plan to run Python MCP servers): curl -LsSf https://astral.sh/uv/install.sh | sh

2. Install git (if not already installed):

   • Ubuntu/Debian: sudo apt-get install git
   • macOS: brew install git

3. Install npx (comes with Node.js): npm install -g npx

4. Install MCPHub (Python package): pip install mcphub

5. Optional: Install framework-specific dependencies for MCPHub: pip install mcphub[openai] # OpenAI integration pip install mcphub[langchain] # LangChain integration pip install mcphub[autogen] # Autogen integration pip install mcphub[all] # All optional dependencies

6. Create and customize your .mcphub.json configuration in your project root as needed (see README examples).

7. Run your MCPHub-enabled workflow or application that loads the MCP servers from the configuration.

Tips and considerations:

• Environment variables can be specified per server in the configuration (e.g., connection strings, API keys). Use the env field to map to ${ENV_VAR} placeholders.
• The configuration supports multiple sources: npm/npx, Python uv, GitHub repos, and local development servers. Combine sources as needed.
• If you use automatic configuration from a GitHub repo, an OpenAI API key must be available via OPENAI_API_KEY.
• Ensure that the server you reference (e.g., smithery-ai/server-sequential-thinking) is compatible with MCPHub's expected tooling interface and exposes tools that OpenAI Agents or LangChain can invoke.
• For debugging, list available tools from a running MCP server to verify proper connection and tool availability before integrating into agents.
• MCPHub handles path management and environment setup, but you may need to adjust setup_scripts in the configuration for non-default installation flows.