This repository contains a complete tutorial on how to create an MCP (Modular Computational Platform) client in Python using FastAPI. The tutorial is inspired by the Claude Desktop application and demonstrates how to make an LLM (Large Language Model) call MCP tools from an MCP server and display the results.
- Build an MCP client using Python and FastAPI.
- Integrate LLMs to interact with MCP tools.
- Fetch and display results from an MCP server.
- Step-by-step guidance for setting up and running the client.
- Python 3.8 or higher
- Basic knowledge of FastAPI and REST APIs
- An MCP server setup
- Clone the Repository:
git clone git@github.com:alejandro-ao/mcp-client-python.git
cd mcp-client-python- Install Dependencies:
pip install -r requirements.txt- Set Up Environment Variables:
Create a
.envfile in the root directory and add your API keys. In this example, I added theSERPER_API_KEYbecause it is used in the MCP server file, which we built in the previous tutorial. And since we are running everything locally, we can set all the environment variables (for both serverand client) in the same file. In a more complex setup, you would deploy your MCP server separately and set the environment variables accordingly.
ANTHROPIC_API_KEY=
SERPER_API_KEY=- Run the Application:
uvicorn main:app --reload- Access the API:
Open your browser and navigate to
http://127.0.0.1:8000/docsto explore the API documentation.
-
Setting Up FastAPI: Learn how to create a FastAPI application and define endpoints.
-
Connecting to the MCP Server: Understand how to make requests to the MCP server and handle responses.
-
Integrating LLMs: Explore how to use an LLM to call MCP tools dynamically.
-
Displaying Results: Implement logic to process and display results from the MCP tools.
- FastAPI Documentation
- MCP Documentation (for inspiration)
This project is licensed under the MIT License. See the LICENSE file for details.
Special thanks to the creators of FastAPI and the Claude Desktop application for their inspiration.