12.5K ·GitHub
What is this?
Chrome MCP Server turns your Chrome browser into a tool AI assistants like Claude can control via the Model Context Protocol (MCP), enabling browser automation, content analysis, and semantic search right inside your everyday browser.
Chrome MCP Server turns your Chrome browser into a tool AI assistants like Claude can control via the Model Context Protocol (MCP), enabling browser automation, content analysis, and semantic search right inside your everyday browser.Is this like a Playwright-based MCP server, or does it use my real Chrome?It is a Chrome extension paired with a local native-host bridge (mcp-chrome-bridge) that hooks into the Chrome browser you already use, instead of launching a separate automated browser instance, so it keeps your existing logins, cookies, and settings.
It is a Chrome extension paired with a local native-host bridge (mcp-chrome-bridge) that hooks into the Chrome browser you already use, instead of launching a separate automated browser instance, so it keeps your existing logins, cookies, and settings.How do I get it running and connect it to my AI client?Download the extension from GitHub, install the mcp-chrome-bridge native host globally via npm/pnpm, load the extension unpacked in chrome://extensions with Developer mode on, then open the extension popup and click connect to get the MCP config (Streamable HTTP is recommended, STDIO is also supported).
Download the extension from GitHub, install the mcp-chrome-bridge native host globally via npm/pnpm, load the extension unpacked in chrome://extensions with Developer mode on, then open the extension popup and click connect to get the MCP config (Streamable HTTP is recommended, STDIO is also supported).What do I need installed before I start?You need Node.js 20 or newer with npm or pnpm, and a Chrome or Chromium-based browser to load the extension into.
You need Node.js 20 or newer with npm or pnpm, and a Chrome or Chromium-based browser to load the extension into.Any rough edges I should know about?The project is still early-stage and under heavy development, and pnpm v7+ disables the bridge's postinstall script by default so you may need to register it manually; Firefox support is only on the roadmap, not available yet.
The project is still early-stage and under heavy development, and pnpm v7+ disables the bridge's postinstall script by default so you may need to register it manually; Firefox support is only on the roadmap, not available yet.Does my browsing data get sent anywhere?No, it runs as a fully local MCP server, so it processes your browser activity on your own machine rather than sending it to a remote service.
No, it runs as a fully local MCP server, so it processes your browser activity on your own machine rather than sending it to a remote service.Why use this instead of a Playwright-based MCP server?Because it runs inside your real, already-open Chrome, it starts almost instantly, reuses your existing login sessions, and gets full access to Chrome's native APIs instead of being limited to Playwright's API surface.
Because it runs inside your real, already-open Chrome, it starts almost instantly, reuses your existing login sessions, and gets full access to Chrome's native APIs instead of being limited to Playwright's API surface.What can the 20+ built-in tools actually do?They cover tab and window management, screenshots, network request capture (including response bodies via the debugger API), page interaction (click, fill, keyboard), content extraction with semantic search, and bookmark/history management.
They cover tab and window management, screenshots, network request capture (including response bodies via the debugger API), page interaction (click, fill, keyboard), content extraction with semantic search, and bookmark/history management.How does the semantic search feature work?It uses a built-in local vector database combined with a custom WebAssembly SIMD module that speeds up vector operations by 4-8x, letting you search across your open tabs by meaning rather than exact keywords.
It uses a built-in local vector database combined with a custom WebAssembly SIMD module that speeds up vector operations by 4-8x, letting you search across your open tabs by meaning rather than exact keywords.What's planned next for the project?The roadmap includes authentication, recording and playback, workflow automation, and extending support beyond Chrome to Firefox.
The roadmap includes authentication, recording and playback, workflow automation, and extending support beyond Chrome to Firefox.What are similar projects?Endoplexity, bb-browser, lean-ctx, XActions, BrowserOS, holaOS, claude-context, mobile-mcp
Endoplexity, bb-browser, lean-ctx, XActions, BrowserOS, holaOS, claude-context, mobile-mcpWhich topics is it in?screenshot, vector-database, semantic-search, chrome-extension, browser-extension, wasm, browser-automation, model-context-protocol, browser, local-first, mcp-server, networking
screenshot, vector-database, semantic-search, chrome-extension, browser-extension, wasm, browser-automation, model-context-protocol, browser, local-first, mcp-server, networking