chrome-devtools-mcp (ChromeDevTools/chrome-devtools-mcp) is the official MCP server from the Chrome DevTools team that exposes browser capabilities to AI agents via the Model Context Protocol. It connects to running Chrome tabs using CDP (Chrome DevTools Protocol), giving agents the ability to inspect DOM elements, read console output, execute JavaScript, capture screenshots, monitor network requests, and interact with page elements. Unlike browser automation tools that simulate user actions at the OS level, this operates at the CDP layer — the same protocol DevTools itself uses. 34K GitHub stars, +243 trending on April 14, 2026.
chrome-devtools-mcp is the Model Context Protocol server maintained by Google’s Chrome DevTools team. It gives AI agents direct access to a running Chrome browser via CDP — the same protocol that powers the DevTools panel you open with F12. The difference from OS-level browser automation is significant: rather than simulating mouse clicks on pixel coordinates, agents send CDP commands that interact with the browser at the protocol layer, with access to the actual DOM tree, JavaScript runtime, and network layer.
With 34K GitHub stars and +243 trending on April 14, 2026, it’s one of the most-starred MCP servers from a first-party source. Being an official Chrome project means it tracks CDP changes with the browser release and has Google engineering support.
DOM inspection: Query the live page structure with full access to element attributes, styles, and computed properties. Agents can find elements by selector, read their content, and understand page layout — not just what’s visible on screen.
JavaScript execution: Run arbitrary JavaScript in the page context. Access window, document, localStorage, and any globals defined by the page. Useful for reading application state that isn’t reflected in the visible UI.
Console and network access: Read all console output (logs, warnings, errors) and monitor network requests and responses. This gives agents observability into application behavior beyond just the visual state.
Screenshot capture: Full-page and element-level screenshots at CDP quality — vector-aware, not rasterized from the screen.
Element interaction: Click, type, scroll, and focus elements via CDP commands. More reliable than coordinate-based clicking because it operates on the actual DOM element reference, not a pixel location that can move.
Connects to any MCP-compatible client: Claude Code (claude mcp add), Cursor, Windsurf, or custom agents using the MCP SDK. Requires Chrome with remote debugging enabled (--remote-debugging-port=9222).
# Start Chrome with debugging
/path/to/Chrome --remote-debugging-port=9222
# Add to Claude Code
claude mcp add chrome-devtools -- npx @chrome-devtools/mcpAI-powered debugging: Give Claude Code access to a running app while you’re developing. The agent can read console errors, inspect network failures, and examine DOM state without you copying output manually.
Browser automation workflows: Agents can navigate, interact with forms, and extract data from live pages as part of a larger workflow — with more reliability than screenshot-based approaches.
Testing assistance: Run automated checks against a live page with full access to application state, not just visible UI.
Open-source AI pair programming tool that works in your terminal to edit code across your entire repository.
AWS's AI-powered coding assistant that helps developers build, deploy, and optimize applications on AWS with code generation and transformation.
Open-source AI coding harness builder that makes AI coding workflows deterministic and repeatable via YAML-defined DAG workflows.