Chrome DevTools MCP (developer.chrome.com)

Google today launched a public preview of the Chrome DevTools Model Context Protocol (MCP) server, an open-source bridge that lets LLM-powered coding assistants interact directly with Chrome DevTools. By exposing DevTools functionality through the MCP standard, AI agents can now open pages in a real browser, inspect DOM and CSS, read console and network logs, simulate user actions, and record performance traces—closing the gap between generated code and observable runtime behavior so agents can debug and verify fixes with real browser evidence. Technically, the server implements MCP tooling (for example, a performance_start_trace tool that launches Chrome, navigates to a page, and captures a trace for LLM analysis), and it integrates into MCP clients via a simple mcpServers entry that runs the chrome-devtools-mcp package (e.g., via npx). Use cases include real-time verification of code changes, diagnosing CORS/network/console errors, reproducing complex user flows, fixing live layout/CSS issues from the live DOM, and automated performance audits (e.g., investigating LCP). The preview is incremental and community-driven—developers and tool vendors are encouraged to try it, file issues, and shape future capabilities via the GitHub repo.