🤖 AI Summary
ChromeDevTools MCP is an open-source Model-Context-Protocol (MCP) server that gives coding agents (Gemini, Claude, Copilot, Cursor, etc.) live, programmatic access to Chrome DevTools. By exposing the browser instance to MCP clients, it lets AI assistants automate navigation with puppeteer, inspect network traffic and console logs, take screenshots, record performance traces, and extract actionable performance insights — enabling reliable automation, deeper debugging, and performance analysis directly from agent workflows.
Technically, chrome-devtools-mcp is launched via npx (recommended: chrome-devtools-mcp@latest) and integrates with many MCP-enabled tools and IDEs; clients add it in their mcpServers config. It provides roughly two dozen specialized tools (navigation, input automation, emulation, performance tracing, network, debugging) and supports options like --headless, --executablePath, --isolated (uses temporary user-data-dir), --browserUrl (connect to an existing remote Chrome), and --channel (stable/canary/beta/dev). Note the default persistent profile location (~/.cache/chrome-devtools-mcp/) unless isolated is used, and a security caveat: MCP clients can read/modify any browser/DevTools data, so avoid exposing sensitive info. Also be aware of sandboxing limits (macOS Seatbelt/Linux containers) which may require disabling sandboxing or connecting to an externally started Chrome.
Loading comments...
login to comment
loading comments...
no comments yet