MCP Starter Project

<p> <img src="https://img.shields.io/badge/License-MIT-yellow?style=flat&colorA=18181B&colorB=28CF8D" alt="License"> <!-- Add other relevant badges here, e.g., build status, stars --> <!-- <a href="YOUR_REPO_LINK/stargazers"><img src="https://img.shields.io/github/stars/YOUR_USERNAME/YOUR_REPO.svg?style=flat&colorA=18181B&colorB=28CF8D" alt="Stars"></a> --> </p>

MCP Starter - A foundation for building your own local Model Context Protocol (MCP) server.

This starter kit provides a basic structure and example setup for creating custom tools accessible via AI assistants like Cursor or Claude Desktop using the MCP standard.

---

Features

• 🚀 MCP Ready - Easily integrate with MCP-compatible clients.
• 🔧 Extensible Tooling - Simple structure to add your own custom tools.
• 📦 Minimal Setup - Get started quickly with a basic server implementation.
• ↔️ Based on Anthropic MCP - Follows the specifications outlined by Anthropic.
• 🤖 Cursor AI Integration - Includes example .cursor/mcp.json configuration.
• ⌨️ TypeScript - Add type safety to your project.

<!-- Add other features specific to your starter implementation -->

Usage

Local development

{
  "mcpServers": {
    "my-starter-mcp": {
      "command": "node",
      "args": ["./dist/index.mjs"]
    }
  }
}

Getting Started

Prerequisites

• Node.js (Specify version if necessary)
• An MCP-compatible client (e.g., Cursor)

Installation & Setup

1. Clone the repository:

   git clone https://github.com/YOUR_USERNAME/YOUR_REPO.git
   cd YOUR_REPO
   

2. Install dependencies:

   npm install
   # or yarn install
   

3. Configure environment variables:

   • Create a .env file based on .env.example.
   • Add any necessary API keys or configuration values required by your custom tools.

4. Add to your MCP client: Add the following MCP configuration to your client (e.g., Cursor's .cursor/mcp.json):

   {
     "mcpServers": {
       "my-custom-mcp": {
         "command": "node", // Or your chosen execution command
         "args": ["path/to/your/server/entrypoint.js"], // Adjust path as needed
         "env": {
           // Add any environment variables your server needs from the client side, if any
           // "EXAMPLE_API_KEY": "<INSERT_API_KEY_HERE>"
         }
       }
     }
   }
   

   • Replace "my-custom-mcp" with a unique name for your server.
   • Adjust the "command" and "args" to correctly point to and run your server's main script.
   • Ensure any required env variables are configured either here or directly in your server's environment (e.g., via the .env file).

Running the Server

npm start
# or yarn start

This command should start your MCP server, making its tools available to connected clients.

Available Tools

• exampleTool: Describe what your example tool does.
• (Add more tools as you implement them)

Develop

This project provides a starting point. You'll likely want to:

• Implement your own custom tools within the server logic.
• Define the schema (parameters, description) for your tools.
• Add error handling and logging.
• Write tests for your tools.

Follow the contribution guidelines if you plan to contribute back to the starter project itself.

License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Links

• X/Twitter: @kregenrek
• Bluesky: @kevinkern.dev

Courses

• Learn Cursor AI: Ultimate Cursor Course
• Learn to build software with AI: instructa.ai

See my other projects:

• AI Prompts - Curated AI Prompts for Cursor AI, Cline, Windsurf and Github Copilot
• codefetch - Turn code into Markdown for LLMs with one simple terminal command
• aidex A CLI tool that provides detailed information about AI language models, helping developers choose the right model for their needs.# mcp-starter