CookiesForAI

Chrome DevTools MCP server

The Chrome DevTools MCP server gives AI agents control of a live Chrome browser with DevTools access: performance traces, network requests, console errors, Lighthouse audits, memory snapshots and page automation.

By Google Chrome DevTools team Official · Updated

Connect Chrome DevTools

Claude Code

Terminal
claude mcp add chrome-devtools --scope user npx chrome-devtools-mcp@latest

Codex

Terminal
codex mcp add chrome-devtools -- npx chrome-devtools-mcp@latest

Gemini CLI

Terminal
gemini mcp add chrome-devtools npx chrome-devtools-mcp@latest

Cursor and other clients with a JSON config

Config
{"mcpServers": {"chrome-devtools": {"command": "npx", "args": ["-y", "chrome-devtools-mcp@latest"]}}}

Add --slim --headless to the args for basic browsing with fewer tools and no window. Performance tools may send page URLs to Google's CrUX API for real-user data; --no-performance-crux turns that off.

What the Chrome DevTools MCP server does

The Chrome DevTools MCP server, which Google calls Chrome DevTools for agents, lets an AI coding agent control and inspect a live Chrome browser. Its edge over plain automation is the DevTools side: the agent can record a performance trace and read the insights out of it, list and open network requests, read console messages with source-mapped stack traces, read computed CSS, and run a Lighthouse audit on a page.

Chrome DevTools MCP also covers the usual page actions through Puppeteer, which waits for each action to finish: navigate, click, fill forms, type, drag, upload files, handle dialogs and emulate devices or slow networks. Recent versions add memory work, where the agent takes heap snapshots and compares them to find leaks, plus screencasts and Chrome extension tools. A slim mode keeps only basic browser tools for agents that need fewer tokens.

The server runs locally with npx and starts Chrome the first time a tool needs it; it can also attach to a Chrome you already have open. The GitHub repo had 52,839 stars on 2026-10-01.

Tools it gives your agent

  • navigate_page
  • click
  • fill_form
  • take_snapshot
  • take_screenshot
  • list_console_messages
  • list_network_requests
  • get_network_request
  • performance_start_trace
  • performance_analyze_insight
  • lighthouse_audit
  • take_heapsnapshot

When to use Chrome DevTools

  • Your page is slow and you want the agent to trace it and name the cause.
  • The agent changed code and should confirm there are no new console errors or failed requests.
  • You want a Lighthouse score and fixes for accessibility or SEO issues.
  • You suspect a memory leak and want heap snapshots compared before and after an action.

When to pick something else

  • Firefox or Safari: Google officially supports only Chrome and Chrome for Testing.
  • Private browsing data: the agent sees everything in the browser it controls, and Google collects usage statistics by default (turn off with --no-usage-statistics).

What Chrome DevTools needs

  • Node.js LTS and npm
  • Google Chrome, current stable or newer

What the Chrome DevTools MCP server costs

Free and open source under Apache-2.0; it drives your own Chrome locally, with no account or usage fees.

Current prices: github.com . We do not list prices: vendors change them often, so check the publisher's own page.

Questions people ask

Is the Chrome DevTools MCP server free?

Yes. Google publishes it under Apache-2.0, and it runs locally against your own Chrome.

Chrome DevTools MCP or Playwright MCP?

Chrome DevTools MCP is stronger for debugging: performance traces, Lighthouse, network detail and memory snapshots, in Chrome only. Playwright MCP covers more browsers and is built around automation and tests.

Does Chrome DevTools MCP send data to Google?

It collects usage statistics such as tool success rates by default, and performance tools can query the CrUX API. Both can be switched off with flags, and statistics stay off when the CI variable is set.

  • Browser Use: The Browser Use MCP server gives AI agents the Browser Use browsing agent, either as a local open-source server with direct browser tools or as a cloud server that runs whole web tasks.
  • Browserbase: The Browserbase MCP server gives AI agents a browser that runs in Browserbase's cloud, controlled with plain-language commands through Stagehand: navigate, act on the page, observe elements and extract data.
  • Playwright: The Playwright MCP server gives AI agents a real browser driven by Microsoft's Playwright: navigate, click, type and fill forms using the page's accessibility tree instead of screenshots.