Have you ever come across a situation where you wanted your chatbot to use a tool and then answer? Sounds complicated, right! But now, MCP (Model Context Protocol)<\/a> offers you a way to integrate your LLM to external tools easily and the LLM will be able to use those tools in every way. In this tutorial, we will dive into the process of converting a simple web app made using FastAPI, powered by an MCP Server, using the FastAPI-MCP.<\/p>\n FastAPI is a very simple tool built in Python which helps you to build web applications using APIs. It is designed to be easy to use as well as fast at the same time. Think of FastAPI as a smart waiter who takes your order (HTTP requests), goes to the Kitchen (Database\/Server) and then takes your order (Output) and then shows it to you. It\u2019s a great tool for building Web backends, Services for Mobile apps etc.<\/p>\n MCP is an open standard protocol by Anthropic that provides a functionality for the LLMs to communicate with external data sources and tools. Think of MCP as a toolkit that provides the right tool for the given task. We\u2019d be using MCP for creating a server.<\/p>\n Now, what if these functionalities are given to your LLM? It will make your life much easier! That\u2019s why FastAPI to MCP integration helps a lot. FastAPI<\/a> takes care of the services from different sources and MCP takes care of the context of your LLM. By using FastAPI with MCP server, we can get access to every tool deployed over the web and utilize that as a LLM tool and make the LLMs do our work more efficiently.<\/p>\n \u00a0In the above image, we can see that there is an MCP server that is connected to an API endpoint. This API endpoint can be a FastAPI endpoint or any other third party API service available on the internet.<\/p>\n FastAPI-MCP is a tool which lets you convert any FastAPI application into some tool that LLMs like ChatGPT<\/a> or Claude can understand and use easily. By using FastAPI-MCP you can wrap your FastAPI endpoints in such a way that they will become a plug and play tool in an AI ecosystem utilizing LLMs.<\/p>\n If you want to know how to work with MCP, read this article on How to Use MCP?<\/a><\/p>\n With FastAPI-MCP, any FastAPI endpoint can be converted into a MCP tool for LLMs. These endpoints should include:<\/p>\n FastAPI-MCP is a very easy-to-use library that automatically discovers and converts these endpoints into MCP. It also preserves the schema as well as the documentation of these APIs.<\/p>\n Let\u2019s look at a simple example on how to convert a FastAPI endpoint into a MCP server. Firstly, we will create a FastAPI endpoint and then move towards converting it into a MCP server <\/a>using Make your system compatible by installing the required dependencies.<\/p>\n Make a new file with the name \u2018main.py\u2019, then import the following dependencies inside it.<\/p>\n Let\u2019s define a FastAPI app with the name \u201cWeather Updates API\u201d.<\/p>\n Now, we will define the routes for our app, which will denote which endpoint will execute which function. Here, we are making a weather update app using weather.gov<\/a> API (free), which doesn\u2019t require any API key. We just need to hit the https:\/\/api.weather.gov\/points\/{lat},{lon}<\/a> with the right value of latitude and longitude.<\/p>\n We defined a get_weather function which will take a state name or code as an argument and then find the corresponding coordinates in the CITY_COORDINATES dictionary and then hit the base URL with those coordinates.<\/p>\n Let\u2019s convert this FastAPI app into MCP now using the fastapi-mcp library. This process is very simple, we just need to add a few lines of and the fastapi-mcp automatically converts the endpoints into MCP tools and detects its schema and documentation easily.<\/p>\n Now, add the following at the end of your Python<\/a> file.<\/p>\n And go to terminal and run the main.py file.<\/p>\n Now your FastAPI app should start in localhost successfully.\u00a0<\/p>\n Let\u2019s configure the Cursor IDE for testing our MCP server.<\/p>\nFastAPI with MCP<\/h2>\n
What is FastAPI-MCP?<\/h2>\n
What APIs Can Be Converted into MCP Using FastAPI-MCP?<\/h2>\n
\n
Hands-on using FastAPI-MCP<\/h2>\n
fastapi-mcp<\/code>.<\/p>\nConfiguring FastAPI<\/h2>\n
1. Install the dependencies<\/h4>\n
pip install fastapi fastapi_mcp uvicorn mcp-proxy<\/code><\/pre>\n2. Import the required dependencies<\/h4>\n
from fastapi import FastAPI, HTTPException, Query\n\nimport httpx\n\nfrom fastapi_mcp import FastApiMCP<\/code><\/pre>\n3. Define the FastAPI App<\/h4>\n
app = FastAPI(title=\"Weather Updates API\")<\/code><\/pre>\n4. Defining the routes and functions<\/h4>\n
# Predefined latitude and longitude for major cities (for simplicity)\n# In a production app, you could use a geocoding service like Nominatim or Google Geocoding API\nCITY_COORDINATES = {\n \"Los Angeles\": {\"lat\": 34.0522, \"lon\": -118.2437},\n \"San Francisco\": {\"lat\": 37.7749, \"lon\": -122.4194},\n \"San Diego\": {\"lat\": 32.7157, \"lon\": -117.1611},\n \"New York\": {\"lat\": 40.7128, \"lon\": -74.0060},\n \"Chicago\": {\"lat\": 41.8781, \"lon\": -87.6298},\n # Add more cities as needed\n}\n\n\n@app.get(\"\/weather\")\nasync def get_weather(\n stateCode: str = Query(..., description=\"State code (e.g., 'CA' for California)\"),\n city: str = Query(..., description=\"City name (e.g., 'Los Angeles')\")\n):\n \"\"\"\n Retrieve today's weather from the National Weather Service API based on city and state\n \"\"\"\n # Get coordinates (latitude, longitude) for the given city\n if city not in CITY_COORDINATES:\n raise HTTPException(\n status_code=404,\n detail=f\"City '{city}' not found in predefined list. Please use another city.\"\n )\n \n coordinates = CITY_COORDINATES[city]\n lat, lon = coordinates[\"lat\"], coordinates[\"lon\"]\n \n # URL for the NWS API Gridpoints endpoint\n base_url = f\"https:\/\/api.weather.gov\/points\/{lat},{lon}\"\n \n try:\n async with httpx.AsyncClient() as client:\n # First, get the gridpoint information for the given location\n gridpoint_response = await client.get(base_url)\n gridpoint_response.raise_for_status()\n gridpoint_data = gridpoint_response.json()\n \n # Retrieve the forecast data using the gridpoint information\n forecast_url = gridpoint_data[\"properties\"][\"forecast\"]\n forecast_response = await client.get(forecast_url)\n forecast_response.raise_for_status()\n forecast_data = forecast_response.json()\n\n\n # Returning today's forecast\n today_weather = forecast_data[\"properties\"][\"periods\"][0]\n return {\n \"city\": city,\n \"state\": stateCode,\n \"date\": today_weather[\"startTime\"],\n \"temperature\": today_weather[\"temperature\"],\n \"temperatureUnit\": today_weather[\"temperatureUnit\"],\n \"forecast\": today_weather[\"detailedForecast\"],\n }\n \n except httpx.HTTPStatusError as e:\n raise HTTPException(\n status_code=e.response.status_code,\n detail=f\"NWS API error: {e.response.text}\"\n )\n except Exception as e:\n raise HTTPException(\n status_code=500,\n detail=f\"Internal server error: {str(e)}\"\n )<\/code><\/pre>\n5. Set up MCP Server<\/h4>\n
mcp = FastApiMCP(\n app,\n name=\"Weather Updates API\",\n description=\"API for retrieving today's weather from weather.gov\",\n)\nmcp.mount()\u00a0<\/code><\/pre>\n6. Starting the app<\/h4>\n
if __name__ == \"__main__\":\n import uvicorn\n uvicorn.run(app, host=\"0.0.0.0\", port=8000)\u00a0<\/code><\/pre>\npython main.py\u00a0<\/code><\/pre>\nConfiguring Cursor<\/h2>\n
\n
![\"Cursor](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/05\/configuring_the_cursor_1.webp\")
<\/figure>\n<\/div>\n\n
![\"Cursor](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/05\/configuring_the_cursor_2.webp\")
<\/figure>\n<\/div>\n\n
![\"Configuring](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/05\/configuring_the_cursor_3-1.webp\")
<\/figure>\n<\/div>\n\n
It will open a mcp.json<\/i> file. Paste the following code into it and save the file.<\/li>\n<\/ol>\n{\n \"mcpServers\": {\n \"National Park Service\": {\n \"command\": \"mcp-proxy\",\n \"args\": [\"http:\/\/127.0.0.1:8000\/mcp\"]\n }\n }\n}<\/code><\/pre>\n\n
![\"Linked](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/05\/Setup_mcp_server2.webp\")
<\/figure>\n<\/div>\n
![\"Prompt](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/05\/testing1.webp\")
<\/figure>\n<\/div>\n![\"Prompt](\"https:\/\/cdn.analyticsvidhya.com\/wp-content\/uploads\/2025\/05\/testing2.webp\")
<\/figure>\n<\/div>\n![\"Harsh](\"https:\/\/av-eks-lekhak.s3.amazonaws.com\/media\/lekhak-profile-images\/converted_image_0fBqNLi.webp\")
<\/p>\n