mcp-chrome

Chrome MCP Server is a Chrome browser extension that lets AI assistants control your Chrome browser directly. It works by implementing the Model Context Protocol (MCP), which is a standard way for AI tools like Claude to communicate with external software. Once the extension is installed and a small bridge program is running on your computer, an AI assistant can open URLs, click things, take screenshots, search your history, manage bookmarks, and more, all through your existing Chrome window.

The key difference from older browser automation tools is that this extension runs inside your real Chrome profile. That means the AI is working in the same browser where you are already signed in to your accounts, and with your saved preferences and extensions in place. Traditional automation tools typically launch a fresh, empty browser instance where you would need to log in again to every site.

The extension includes over 20 tools covering browser management, screenshots, page interaction, bookmarks, history, and a semantic search feature that uses a vector database to find relevant content across your open tabs. The semantic search is accelerated using a custom WebAssembly SIMD optimization that the README says produces a 4x to 8x speed improvement for the underlying vector operations.

Setup requires installing the Chrome extension manually in developer mode, then installing a companion npm package called mcp-chrome-bridge. Once both are running, you point your MCP-compatible AI client at a local HTTP address and the browser tools become available to it.

The project is MIT licensed and still described as being in early-stage development.