Multi MCP is a Python-based server that acts as a proxy to multiple MCP servers. It exposes an API surface and transport options (SSE or streamable HTTP) so clients can interact with several MCP backends through a single gateway. The server reads its configuration from a MCP tools JSON file and can be started with a standard ASGI-compatible runner. Once running, you can access the UI or endpoints that expose the proxied MCP tools, making it easier to manage multiple MCP endpoints from one place.

Prerequisites:

• Python 3.8+ (recommended) and pip
• Git
• Optional: a Python virtual environment tool (e.g., venv, virtualenv)

Install and run:

1. Clone the repository: git clone https://github.com/One-MCP/multi-mcp.git cd multi-mcp

2. Create and activate a virtual environment (optional but recommended): python -m venv venv

   On Windows:

   venv\Scripts\activate

   On macOS/Linux:

   source venv/bin/activate

3. Install dependencies: uv pip install -r requirements.txt

4. Start the server (as per the project README): hypercorn src.main:app

   Alternatively, with explicit module execution:

   python -m hypercorn src.main:app

5. Ensure configuration is available via CONFIG path or defaults, and start using the endpoints documented in the README.

Environment variables determine authentication, configuration source, and transport behavior. Common variables to set include AUTH_TOKEN (optional), CONFIG (path to MCP configuration JSON), ENCODING, TRANSPORT (stdio, sse, or http), USE_NAMESPACE, and DEBUG. If you encounter authentication issues, verify the token format (Bearer token for Authorization header or plain token for X-MCP-Token). The server can proxy multiple MCP endpoints defined in the configuration file; ensure each target MCP server is reachable from the host running this gateway. For development, you can switch TRANSPORT to http to test via standard HTTP requests, or sse for server-sent events streaming.