Chrome DevTools MCP is the official Model Context Protocol server from the Chrome team, with 53,900+ stars on GitHub, that gives AI coding agents eyes and hands inside a real Chrome browser.
Coding agents can edit your code, but they are blind to what the browser actually does with it. Chrome DevTools MCP closes that gap: it connects agents such as Claude Code, Cursor, Copilot and Gemini CLI to the full DevTools feature set through MCP. The agent can launch or attach to Chrome, navigate pages, click and fill forms, read console messages and network requests, inspect the DOM and styles, take screenshots, and record performance traces with the same insights a developer would get from the Performance panel. It is built by the Chrome DevTools team on top of Puppeteer, distributed as an npm package, and documented as the reference way to give an assistant real debugging access to a web app. The project is under very active development, with updates landing in October 2026.
Chrome DevTools MCP is free and open source under the Apache License 2.0, published by the Chrome team. There is no paid tier; you only need Chrome and an MCP-capable agent to use it.
Web developers and QA engineers who work with AI coding agents and want those agents to debug layout, network and performance problems directly in Chrome instead of working blind from source code alone.