Practical MCP with FastMCP & LangChain

Running the Server

The server is a uvicorn ASGI application. Navigate to the server/ directory and start it with:

.venv/bin/python -m uvicorn main:app --reload --port 8000

The --reload flag makes uvicorn watch for file changes and restart automatically, which is useful during development. We use uvicorn directly rather than fastmcp run --reload because the latter forces stateless mode: it generates a fresh session ID on every request, which means the Redis-backed favorites list would be lost between calls. Running through uvicorn preserves the mcp-session-id header and keeps session state alive across the whole conversation.

For reference, the server directory also contains a fastmcp.json configuration file. This file describes how to run the server with fastmcp run and is useful if you want to deploy to a managed environment in the future.

We're not using it but here's what it looks like:

{
  "$schema": "https://gofastmcp.com/public/schemas/fastmcp.json/v1.json",
  "source": {
    "path": "main.py",
    "entrypoint": "mcp"
  },
  "environment"