mcp-hub

MCP Hub is a framework for creating and managing Model Context Protocol (MCP) servers and clients. It leverages the UV tool for fast package installation and configuration management, enabling you to scaffold projects, manage dependencies, and run MCP servers with simple commands. With MCP Hub you can create new MCP-compliant servers, wire up client interactions, and use UV to install and keep dependencies up to date. Typical workflows involve initializing a project, setting up a Python virtual environment, adding MCP-related CLI tooling, and executing your MCP server script for testing and deployment. The included examples in the documentation show a straightforward path from project creation to running a server with UV.

Prerequisites:

• Python 3.8+ (for MCP server development and execution)
• Git
• Curl or Wget (for UV installation script)
• A Unix-like environment (Linux/macOS) or Windows Subsystem for Linux (WSL) if you are on Windows

Installation steps:

1. Install UV (the fast package manager used by MCP Hub):

   • Unix/macOS/Linux: curl -fsSL https://uv.sh/install.sh | bash
   • Follow the on-screen instructions to ensure the 'uv' CLI is available in your PATH.

2. Create a new project or clone an existing MCP Hub template

   • Initialize a new project (example name: my-mcp-project): uv init my-mcp-project
   • Change into the project directory: cd my-mcp-project

3. Set up a Python virtual environment (recommended):

   • uv venv
   • source .venv/bin/activate

4. Install MCP Hub CLI and dependencies for MCP development:

   • uv add "mcp[cli]" httpx

5. Create or place your MCP server script (e.g., my_server.py) and ensure it can be executed by uv run:

   • uv run my_server.py

Optional for faster iteration with a sample server:

• If you are using the template for the MCP Hub, you can customize the server script and configuration files, then run the server using UV as shown above.

Tips and common issues:

• Ensure your Python virtual environment is activated before installing or running MCP dependencies.
• If UV cannot locate your package name (as in a private repository), verify your UV configuration and authentication settings for private registries.
• For Windows users, consider using WSL to match a Unix-like environment for UV and MCP Hub workflows.
• When starting new servers, specify a clear server name in the mcp_config to track multiple MCP servers easily.
• Environment variables such as MCP_HOST, MCP_PORT, or API keys may be required by your specific MCP server implementation; add them to your environment or an .env file as appropriate.
• If you encounter networking or dependency resolution issues with UV, run uv sync to synchronize Python versions and dependencies as described in the UV guide.