π Turn your Chrome browser into your intelligent assistant - Let AI take control of your browser, transforming it into a powerful AI-controlled automation tool.
π Documentation: English | δΈζ
The project is still in its early stages and is under intensive development. More features, stability improvements, and other enhancements will follow.
Chrome MCP Server is a Chrome extension-based Model Context Protocol (MCP) server that exposes your Chrome browser functionality to AI assistants like Claude, enabling complex browser automation, content analysis, and semantic search. Unlike traditional browser automation tools (like Playwright), Chrome MCP Server directly uses your daily Chrome browser, leveraging existing user habits, configurations, and login states, allowing various large models or chatbots to take control of your browser and truly become your everyday assistant.
- A New Visual Editor for Claude Code & Codex, for more detail here: VisualEditor
- π Chatbot/Model Agnostic: Let any LLM or chatbot client or agent you prefer automate your browser
- βοΈ Use Your Original Browser: Seamlessly integrate with your existing browser environment (your configurations, login states, etc.)
- π» Fully Local: Pure local MCP server ensuring user privacy
- π Streamable HTTP: Streamable HTTP connection method
- π Cross-Tab: Cross-tab context
- π§ Semantic Search: Built-in vector database for intelligent browser tab content discovery
- π Smart Content Analysis: AI-powered text extraction and similarity matching
- π 20+ Tools: Support for screenshots, network monitoring, interactive operations, bookmark management, browsing history, and 20+ other tools
- π SIMD-Accelerated AI: Custom WebAssembly SIMD optimization for 4-8x faster vector operations
| Comparison Dimension | Playwright-based MCP Server | Chrome Extension-based MCP Server |
|---|---|---|
| Resource Usage | β Requires launching independent browser process, installing Playwright dependencies, downloading browser binaries, etc. | β No need to launch independent browser process, directly utilizes user's already open Chrome browser |
| User Session Reuse | β Requires re-login | β Automatically uses existing login state |
| Browser Environment | β Clean environment lacks user settings | β Fully preserves user environment |
| API Access | β Full access to Chrome native APIs | |
| Startup Speed | β Requires launching browser process | β Only needs to activate extension |
| Response Speed | 50-200ms inter-process communication | β Faster |
Self-build install: no npm install -g, no release zip β you build the extension and native bridge yourself, then register them together.
- Node.js >= 20 (tested on v22)
pnpm(npm install -g pnpmif you don't have it)- Chrome / Chromium
Rust toolchain is not required β the installer reuses the prebuilt WASM module shipped in
releases/.
git clone https://github.com/itorz7/mcp-chrome.git
cd mcp-chrome
./scripts/install.sh buildThis installs workspace dependencies, copies the prebuilt WASM, then builds the shared package, the native bridge, and the Chrome extension. When it finishes you'll see two paths:
app/chrome-extension/.output/chrome-mv3β the unpacked extensionapp/native-server/distβ the native messaging bridge
- Open
chrome://extensions - Toggle Developer mode on
- Click Load unpacked and select
app/chrome-extension/.output/chrome-mv3 - Copy the extension ID Chrome assigns (shown right under the extension name)
./scripts/install.sh register <YOUR_EXTENSION_ID>Chrome generates a fresh extension ID every time the unpacked folder moves. Each time you move the repo or re-load the extension you'll need to re-run
registerwith the new ID. The popup shows the current ID and the exact command to run β just copy.
Chrome caches native messaging manifests when it launches, so after registering you must Cmd+Q / quit all Chrome windows and relaunch. Then open the extension popup and click Connect.
The bridge exposes Streamable HTTP on port 12306:
{
"mcpServers": {
"chrome-mcp": {
"type": "http",
"url": "http://127.0.0.1:12306/mcp"
}
}
}Claude Code:
claude mcp add --scope user --transport http chrome-mcp http://127.0.0.1:12306/mcp./scripts/install.sh build # rebuild bridge + extension
./scripts/install.sh register <EXT_ID> # re-register (only if ID changed)
./scripts/install.sh unregister # wipe the native host manifest- "Native host has exited" in service worker console β the
allowed_originsin the manifest does not match the extension ID. Re-runregisterwith the current ID fromchrome://extensions. - "Connected, Service Not Started" β the popup connected to the native host but the HTTP server hasn't started. Click Disconnect then Connect again, or fully quit Chrome first.
- Manifest not picked up after
registerβ Chrome caches manifests on launch. Quit Chrome completely (Cmd+Qon macOS) and reopen. - macOS and project lives under
~/Desktopβ TCC can block Chrome from spawning the wrapper script. Move the repo out of~/Desktop(e.g. to~/mcp-chrome) and re-runregister.
Complete tool list: Complete Tool List
π Browser Management (6 tools)
get_windows_and_tabs- List all browser windows and tabschrome_navigate- Navigate to URLs and control viewportchrome_switch_tab- Switch the current active tabchrome_close_tabs- Close specific tabs or windowschrome_go_back_or_forward- Browser navigation controlchrome_inject_script- Inject content scripts into web pageschrome_send_command_to_inject_script- Send commands to injected content scripts
πΈ Screenshots & Visual (1 tool)
chrome_screenshot- Advanced screenshot capture with element targeting, full-page support, and custom dimensions
π Network Monitoring (4 tools)
chrome_network_capture_start/stop- webRequest API network capturechrome_network_debugger_start/stop- Debugger API with response bodieschrome_network_request- Send custom HTTP requests
π Content Analysis (4 tools)
search_tabs_content- AI-powered semantic search across browser tabschrome_get_web_content- Extract HTML/text content from pageschrome_get_interactive_elements- Find clickable elementschrome_console- Capture and retrieve console output from browser tabs
π― Interaction (3 tools)
chrome_click_element- Click elements using CSS selectorschrome_fill_or_select- Fill forms and select optionschrome_keyboard- Simulate keyboard input and shortcuts
π Data Management (5 tools)
chrome_history- Search browser history with time filterschrome_bookmark_search- Find bookmarks by keywordschrome_bookmark_add- Add new bookmarks with folder supportchrome_bookmark_delete- Delete bookmarks
prompt: excalidraw-prompt Instruction: Help me summarize the current page content, then draw a diagram to aid my understanding. https://www.youtube.com/watch?v=3fBPdUBWVz0
use.chrome-mcp.to.summarize.web.content.and.then.draw.with.Excalidraw.mp4
After analyzing the content of the image, the LLM automatically controls Excalidraw to replicate the image
prompt: excalidraw-prompt|content-analize Instruction: First, analyze the content of the image, and then replicate the image by combining the analysis with the content of the image. https://www.youtube.com/watch?v=tEPdHZBzbZk
chrome-mcpa.ae.a.a.c.a.a.a.c.a.mp4
prompt: modify-web-prompt Instruction: Help me modify the current page's style and remove advertisements. https://youtu.be/twI6apRKHsk
use.chrome-mcp.to.modify.web.style.mp4
query: I want to know what the search API for Xiaohongshu is and what the response structure looks like
1hHKr7XKqnQ.mp4
query: Analyze my browsing history from the past month
jf2UZfrR2Vk.mp4
query: Translate and summarize the current web page https://youtu.be/FlJKS9UQyC8
FlJKS9UQyC8.mp4
query: Take a screenshot of Hugging Face's homepage https://youtu.be/7ycK6iksWi4
7ycK6iksWi4.mp4
query: Capture the icon from Hugging Face's homepage https://youtu.be/ev8VivANIrk
ev8VivANIrk.mp4
query: Add the current page to bookmarks and put it in an appropriate folder
R_83arKmFTo.mp4
query: Close all shadcn-related web pages
2wzUT6eNVg4.mp4
We welcome contributions! Please see CONTRIBUTING.md for detailed guidelines.
We have exciting plans for the future development of Chrome MCP Server:
- Authentication
- Recording and Playback
- Workflow Automation
- Enhanced Browser Support (Firefox Extension)
Want to contribute to any of these features? Check out our Contributing Guide and join our development community!
This project is licensed under the MIT License - see the LICENSE file for details.
- Architecture Design - Detailed technical architecture documentation
- TOOLS API - Complete tool API documentation
- Troubleshooting - Common issue solutions