MCP Puppeteer

🚀 A Simple MCP Server for Controlling Chrome via Puppeteer

This project provides a straightforward MCP server that enables you to control Chrome using Puppeteer. It's a practical solution for those looking to integrate browser automation into their workflows.

🚀 Quick Start

⚠️ Important Note

Note that this project uses puppeteer-core, which won't install an additional browser. It's expected that you know how to start Chrome with the remote debugging port enabled. For example, on a Mac, it might look like this:

open -a "Google Chrome" --args --remote-debugging-port=9222

📦 Installation

1. Install the dependencies:

% pnpm i

2. Run the MCP server:

% pnpm run start

3. In Cursor → Settings → Cursor Settings → MCP → Add a new global MCP server, add the following configuration:

{
  "mcpServers": {
    "puppeteer-control": {
      "url": "http://localhost:7742/sse"
    }
  }
}

📚 Documentation

💡 Learning Points

1. The TypeScript SDK documentation describes the newer "Streamable HTTP" support, which is demonstrated in its examples. However, for Cursor, we need SSE (Server-Sent Events) support, which is explained in more detail in the official MCP documentation.
2. The Inspector is very useful, but its usage might be a bit misleading. Assuming the local MCP server is running and using SSE, you can simply use the pnpx @modelcontextprotocol/inspector command to start it.