Skip to content
ML2

Chrome DevTools MCP: let your coding agent debug in real Chrome

Added NewCopyable prompt below

Published Last verified Released

Our verdict

The fix for agents that say "fixed" without ever loading the page: with this server the agent can reproduce the bug in a real browser, read the console error and failing request, change the code and check again. It has been public since September 2025, ships often (v1.13.0 on 11 Nov 2026), sits at about 57,000 GitHub stars and was on GitHub Trending again this week. Since Chrome 144 it can also attach to the Chrome window you are already using (--autoConnect), so the agent can debug a page behind your sign-in or a request you selected in the Network panel; Chrome asks for permission each time. On Hacker News (604 points), boomskats reports using it daily with Codex in opencode and finding it more reliable and token-efficient than other DevTools-protocol MCP servers, including for Electron apps. Bundled skills cover accessibility audits, cookie and session debugging, LCP fixes and memory leaks. Limits: the agent can see and change anything in that browser, so keep sensitive accounts out of the profile it uses; Google collects usage statistics by default (--no-usage-statistics to opt out) and performance tools may send trace URLs to the CrUX API (--no-performance-crux); only Chrome and Chrome for Testing are officially supported. For scripted end-to-end flows across browsers, Playwright MCP is still the better fit. We have not run it ourselves.

Who it's for
Web developers using a coding agent who want it to verify UI, network and performance fixes in a real browser instead of guessing from source
Leverage
Your agent checks its own front-end work against real console, network and performance data, so fewer "fixed" claims turn out to be wrong and slow-page fixes start from a trace, not a hunch
Effort
Low
Date added

What it does

chrome-devtools-mcp (ChromeDevTools/chrome-devtools-mcp, Apache-2.0) runs with npx and uses Puppeteer to drive Chrome, waiting for each action's result. Its tool reference lists input automation (click, fill, navigate, pages), emulation, performance traces with extracted insights, network request inspection, screenshots and console reading. The browser starts only when a tool needs it, using a dedicated profile by default; options include --headless, --isolated (temporary profile), --slim (a reduced tool set for basic browsing), connecting to a running Chrome over a remote debugging port, or --autoConnect to an active session in Chrome 144 and later. The same package installs a chrome-devtools CLI for shell scripts and seven skills; plugins for Claude Code, Copilot in VS Code and Gemini CLI install the server and skills together. It needs current Node.js LTS and current stable Chrome, checks npm for updates (CHROME_DEVTOOLS_MCP_NO_UPDATE_CHECKS to disable), and skips usage statistics when CI is set.

How builders use it

  • Add the server to your agent (one command for Claude Code or Codex, or the JSON block for other MCP clients), or install the Claude Code plugin to get the skills too.
  • Test it with "Check the performance of https://developers.chrome.com" and confirm a browser opens and a trace is recorded.
  • When a UI bug is reported, ask the agent to reproduce it on your local dev server, read the console and network panel, fix it and reload to confirm before it reports back.
  • For bugs behind a sign-in, enable remote debugging at chrome://inspect#remote-debugging, run the server with --autoConnect and approve Chrome's prompt.

What to expect

Front-end fixes arrive already checked in a browser, and performance work starts from real traces and LCP data rather than the agent reading source code.

Copy this: Setup + verify-before-done prompt

# Official source: github.com/ChromeDevTools/chrome-devtools-mcp (Apache-2.0)
# Claude Code:
claude mcp add chrome-devtools --scope user npx chrome-devtools-mcp@latest
# Codex:
codex mcp add chrome-devtools -- npx chrome-devtools-mcp@latest
# Other MCP clients:
{ "mcpServers": { "chrome-devtools": { "command": "npx", "args": ["-y", "chrome-devtools-mcp@latest", "--no-usage-statistics"] } } }
# Plugin with all skills (Claude Code):
/plugin marketplace add ChromeDevTools/chrome-devtools-mcp
/plugin install chrome-devtools-mcp@chrome-devtools-plugins

# Prompt:
Open http://localhost:3000/<page> in Chrome. Reproduce the bug: <steps>. Read the console messages and failed network requests, find the cause in the code and fix it. Then reload, repeat the steps and show me the console and a screenshot proving it works. Do not say it is fixed until you have checked in the browser.

Related discoveries

See the newest picks →
  • ToolSmall Tools, Repos & Plugins

    agent-device: verify mobile apps from your coding agent

    CLI, MCP server and Node API that let coding agents inspect, control and verify iOS, Android and other apps with accessibility snapshots and evidence.

    Effort: MediumPrompt#Plugin / extension#MCP

  • ToolCoding & Shipping Agents

    goose

    An open-source agent with a desktop app, CLI and API that runs locally and works with 15+ model providers.

    Effort: Low#Local / BYO model#MCP

  • ToolCoding & Shipping Agents

    Sphinx: quiz yourself before shipping agent code

    MIT Git hook that quizzes you on what your coding agent's change does before a protected push continues.

    Effort: MediumPrompt#Plugin / extension#Open source

Get the best new A.I. picks in your inbox

The best new agent workflows and tools we’ve vetted, with a clear verdict on each. Short, occasional, useful.

More in Coding & Shipping Agents →

Free. No spam. Unsubscribe anytime.