Creating a Python MCP Server: A Step-by-Step Guide

December 30, 2024 (1y ago) · 2 min read

MCP

Originally published on LinkedIn, 30 December 2024. Republished here with light editing.

Introduction#

This guide will walk you through the process of creating a Python MCP (Model Control Protocol) Server from scratch using the uvx tool. We'll cover everything from initial setup to project structure.

Prerequisites#

  • Python 3.12 or higher installed
  • Windows PowerShell or Command Prompt
  • Basic understanding of Python

Step 1: Installation and Project Creation#

First, navigate to your desired project directory:

cd "D:\github_repository"

Create a new MCP server project using the uvx command:

uvx create-mcp-server

During the creation process, you'll be prompted for:

  • Project name (we used "server-google-jobs")
  • Project description (optional)
  • Project version (default: 0.1.0)
  • Confirmation of project location

Step 2: Virtual Environment and Dependencies#

The tool automatically:

  1. Creates a virtual environment in .venv directory
  2. Installs necessary dependencies including: mcp==1.1.2 pydantic==2.10.4 starlette==0.45.0 httpx==0.28.1 And other required packages

To ensure all dependencies are properly installed, run:

cd server-google-jobs
uv sync --dev --all-extras

Step 3: Project Structure#

The generated project structure looks like this:

SERVER-GOOGLE-JOBS/
├── .venv/
├── Lib/
│   └── site-packages/
├── Scripts/
├── src/
│   └── server_google_jobs/
│       ├── __pycache__/
│       │   ├── __init__.cpython-312.pyc
│       │   └── server.cpython-312.pyc
│       ├── __init__.py
│       └── server.py
├── .gitignore
├── CACHEDIR.TAG
├── pyproject.toml
├── pyenv.cfg
├── .python-version
├── README.md
└── uv.lock

Key files and their purposes:

  • src/server_google_jobs/server.py: Main server implementation file
  • src/server_google_jobs/__init__.py: Package initialization
  • pyproject.toml: Project configuration and dependencies
  • .gitignore: Specifies which files Git should ignore
  • README.md: Project documentation

Notes#

  • The server is automatically configured for Claude.app integration if detected
  • The configuration file is stored at %APPDATA%\Claude\claude_desktop_config.json
  • The project uses Python 3.12.2 by default

Next Steps#

After setting up the project structure, you can:

  1. Implement your server logic in server.py
  2. Add custom routes and endpoints
  3. Configure authentication if needed
  4. Add tests and documentation
  5. Deploy your server

Common Issues#

If you encounter the hardlink warning during installation:

warning: Failed to hardlink files; falling back to full copy

This is normal and won't affect functionality. You can suppress it by setting:

export UV_LINK_MODE=copy

or using the --link-mode=copy flag.

Conclusion#

You now have a fully configured Python MCP Server project ready for development. The structure provided gives you a solid foundation to build upon, with all necessary dependencies and configurations in place.

LinkedIn