mcp-server

mcp-server

3.4

InBrowserMCP Project enables AI model interaction with a web browser through a browser extension using the Model Context Protocol (MCP).

The InBrowserMCP Project is designed to facilitate interaction between an AI model and a web browser using a browser extension. It comprises three main components: a React-based frontend for user interaction, a Node.js MCP server that manages communication and command interpretation, and a Chrome browser extension that executes browser actions. The architecture allows for seamless communication between these components, enabling the execution of browser commands such as navigation, clicking, and content retrieval. The frontend sends user or AI commands to the MCP server, which interprets and translates them into actions for the browser extension. The extension performs these actions and returns results to the server, which then updates the frontend. This setup allows for dynamic browser control and interaction through a structured protocol.

Features

  • Seamless integration between frontend, server, and browser extension.
  • Real-time browser control via WebSocket communication.
  • Server-Sent Events (SSE) for streaming updates to the frontend.
  • React-based user interface for intuitive interaction.
  • Node.js backend implementing the MCP specification.

Usage with Different Platforms

mcp_server

bash
cd mcp-server && pnpm dev

browser_extension

bash
Load the extension in Chrome via chrome://extensions/

frontend

bash
cd frontend && pnpm dev