falcon

falcon-mcp-server exposes a Falcon MCP server that can be run as an ASGI/HTTP-based MCP endpoint and interacted with by clients speaking the Model Context Protocol (Streamable HTTP transport). The project provides a minimal server you can deploy locally or in a container, with guidance in the repository on running an example app via Uvicorn. You can connect clients such as MCP inspectors or agents (e.g., MCPhost, Ollama) to this server to exchange structured, streamable HTTP messages defined by the MCP specification. The example demonstrates creating a simple app with an MCP factory, then running the server and using a compatible transport to issue queries and stream responses. In practice, you’ll start the server, then connect a client capable of MCP transport to issue prompts or requests and receive streaming results.

Prerequisites:

• Python 3.8+ (or whichever version is supported by the falcon-mcp-server project)
• pip

Installation steps:

1. Install the package from PyPI with the extra dependencies for serving:

   pip install falcon-mcp-server[serve]
   

2. (Optional) If you prefer to install directly from the main branch for the latest changes:

   pip install git+https://github.com/falconry/falcon-mcp-server
   

3. Verify installation by listing package version (optional):

   python -m pip show falcon-mcp-server
   

4. Run the example app as shown in the repository (see next section for exact command patterns).

Tips and notes:

• The server is designed to work with the Streamable HTTP transport version of MCP. Ensure your client and server agree on the transport version.
• The example usage shows running an app via Uvicorn. Depending on your deployment, you can adapt to your preferred ASGI server or run the app directly if a module entry point is provided by the package.
• If you need interactive usage, you can omit the prompt for a chat-like experience; otherwise, pass a prompt to drive the MCP interaction.
• Using an inspector like MCPHost or Ollama requires compatible configuration and model/model-provider access; consult their docs for integration specifics.
• For local development, you may mount an example directory as shown in the repository and run uvicorn with the factory function to create your app.