Build an Anthropic MCP Server with FastAPI in Minutes

Anthropic's MCP protocol enables seamless, clear connections between AI agents and your applications, and implementing it with Python is straightforward.

Why MCP?

Allows AI agents to integrate with your application.

Shifts complexity from human developers to AI agents, potentially simplifying development.

Simplifies connecting AI to many tools and data sources.

If MCP becomes the next standard, adopting it will be essential as developers move from manual integration to AI‑driven automation.

How to Build

Create a FastAPI Server

(If you already have a FastAPI server, you can skip this section.)

Core steps:

Install dependencies:

<code>pip install uvicorn fastapi</code>

Create a FastAPI server:

<code>from fastapi import FastAPI

app = FastAPI()

@app.get("/")
async def root():
    return {"message": "MCP is really cool"}
</code>

Write a path operation for each feature (the root function is bound to the path / and handles HTTP GET requests).

Run your application with uvicorn:

<code>uvicorn main:app --reload</code>

Now your FastAPI app is reachable at http://127.0.0.1:8000 . Next, upgrade it to an MCP server.

From FastAPI to MCP Server

Install the open‑source fastapi-mcp tool:

<code>pip install fastapi-mcp</code>

Add the following code to your FastAPI application:

<code>from fastapi import FastAPI
from fastapi_mcp import add_mcp_server

app = FastAPI()
add_mcp_server(
    app,
    mount_path="/mcp",
    name="My API MCP",
)
</code>

That’s it! The MCP server is generated automatically and can be accessed at http://127.0.0.1:8000/mcp . This address can be configured in Cursor (or any AI agent supporting SSE); support in Cline is rumored.

Using MCP

Configure your MCP in Cursor:

Settings → MCP → Add new MCP.

Paste the following JSON into the opened file:

<code>{
  "mcpServers": {
    "My First MCP server": {
      "url": "http://127.0.0.1:8000/mcp"
    }
  }
}
</code>

Claude Desktop requires an extra step because it currently only supports stdio; a future article will cover that.

What Next?

After configuration, AI agents can interact with your application via the standardized protocol. Expand your MCP server by adding more endpoints to expose additional functionality.

Remember, MCP aims to simplify integration—start with basic features and grow as needed.

Using FastAPI together with the fastapi-mcp library retains all FastAPI advantages (auto‑generated docs, type validation) while adding MCP capabilities with minimal code.