MCP Server Kickstarter

Create a New Project

Use the uv CLI to initialize a new project:

uv init <<project-name>>

This command creates a new MCP-compatible server project in the current directory.

Create and Activate a Virtual Environment

Create a virtual environment for the project:

uv venv

Activate it with:

source .venv/bin/activate

Add Dependencies

Use the uv CLI to add dependencies to your project. For example, add LangChain with:

uv add langchain

Install project dependencies from the lock file:

uv sync

Create or update the dependency lock file:

uv lock

Run the main Python file for your project:

uv run main.py

Debug Inspector

Use the MCP inspector CLI for debugging your server:

npx @modelcontextprotocol/inspector

Inspector Configuration

Transport Type: stdio

Command: uv

Arguments:

run --directory /path-to-directory/mcp-server-kickstarter main.py

Debugging Steps

1. Start the Inspector

   npx @modelcontextprotocol/inspector uv run main.py
   

2. Click "Connect"

   • In the inspector web interface, click the Connect button
   • This establishes a connection to your MCP server

3. List Available Tools

   • Click the "List Tools" button in the inspector
   • This displays all tools exposed by your server in JSON format
   • Each tool shows its name, description, and input schema

4. Test Tools

   • Select a tool from the list
   • Enter the required parameters
   • Click "Call Tool" to execute it
   • View the response in the output panel

Register Server in VS Code

To connect your MCP server in VS Code, add it to your mcp.json file:

Workspace or Project Configuration

Add the following server configuration to your mcp.json:

{
  "servers": {
    "get_local_document": {
      "type": "stdio",
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "project-folder/mcp-server-kickstarter",
        "fastmcp_mock_db.py"
      ]
    }
  }
}

Configuration Breakdown

• type: stdio - Uses standard input/output transport
• command: uv - Uses the uv package manager
• args:
  • run - Executes the Python file
  • --directory - Specifies the project directory path
  • project-folder/mcp-server-kickstarter - Path to your MCP server folder
  • fastmcp_mock_db.py - The server file to run

After adding this configuration, the server will be available in VS Code's MCP tools and can be used in Copilot.