Browser MCP over CDP

What is Browser MCP over CDP?

Browser MCP over CDP is an MCP server that manages a Chrome browser using the Chrome DevTools Protocol (CDP). It exposes tools to navigate, interact with, and access a browser tab, exporting the page as a full-height screenshot, accessibility tree, and enriched deep DOM. It is intended for developers who need to programmatically control a browser from MCP clients like Claude, VS Code, or Cursor.

How to use Browser MCP over CDP?

Prerequisites are Docker, Docker Compose, and Python. Run Selenoid Selenium Hub by pulling the required Docker images and executing docker-compose up inside the selenoid directory. Copy and adjust config-example.json, then configure it for your MCP client (Claude, VS Code, or Cursor). View the live browser at http://localhost:8080/. A playground notebook (playground.ipynb) is also provided.

Key features of Browser MCP over CDP

• Full-page screenshot capture (not just viewport)
• Accessibility tree extraction with relevant parts
• Enriched deep DOM including iframes, shadow DOM, and listeners
• Interactive actions: click, keys, input value updates
• Navigation: open URL, back, forward, reload
• Remote browser via Docker with Selenoid and VNC

Use cases of Browser MCP over CDP

• Automate browser tasks from MCP clients like Claude or Cursor
• Capture full-page screenshots for visual analysis or archiving
• Interact with complex web pages programmatically via clicks and keyboard
• Extract enriched DOM and accessibility data for AI‑powered web understanding
• Navigate browser history and reload pages during automation