Search skills/

Chrome DevTools MCP Claude skills

Google's Chrome DevTools MCP server and CLI, plus seven skills for debugging performance, memory leaks, cookies and accessibility in a live Chrome.

ChromeDevTools/chrome-devtools-mcp 7 skills Planning, TDD & debuggingBrowser automation & scraping
Our verdict Recommended with caveats

A first-party, Apache-2.0 browser debugging toolkit with solid skills, but usage statistics are on by default and the agent gets full access to the browser it controls.

3/5 Depth expertise shipped Note Security Review before installing 0d Maintenance Updated today 54K Adoption stars OK License Apache-2.0, commercial OK

What Chrome DevTools MCP does

This repo is the official Chrome DevTools MCP server, an npm package that lets a coding agent control and inspect a live Chrome through Puppeteer and DevTools. The agent can navigate pages, click, read the console with source-mapped stack traces, inspect network requests, take screenshots, evaluate scripts, record performance traces and run Lighthouse audits. The same package includes a chrome-devtools CLI so you can drive the browser from a terminal without MCP.

The seven skills tell the agent how to use those tools for specific jobs. a11y-debugging checks accessibility issues, debug-optimize-lcp breaks Largest Contentful Paint into subparts and points at the biggest delay, memory-leak-debugging uses heap snapshots (it needs the server started with --memoryDebugging), cookie-debugging explains HttpOnly versus client-side cookies and when to use a live tab or an isolated context, troubleshooting helps when the server will not connect and has the agent read your MCP config files, and two skills cover the server and the CLI. SKILL.md files average about 7 KB with a handful of reference files and no scripts. The README says it officially supports Google Chrome and Chrome for Testing, and it warns that connected clients can see and modify browser content.

Good for

  • Debugging a web app's performance, memory, network and cookies from an agent
  • Running Lighthouse and trace analysis against a real Chrome
  • Teams that want a first-party Google tool rather than a third-party browser MCP

Not for

  • Browsers other than Chrome (support is not guaranteed)
  • Sessions where the browser holds personal or sensitive data you do not want an agent to see

What's inside depth 3/5

Seven skills averaging 7 KB with a few reference files and no scripts, backed by a full MCP server and CLI that carry most of the capability.

Skills
7
Reference docs
7
Scripts
0
Tests / evals
0
Templates
0
Avg SKILL.md
7.1 KB

Counted from the repository file tree. How we score depth.

Dependencies and costs

  • Node.js and npm (the server runs via npx chrome-devtools-mcp@latest)
  • Current stable Google Chrome or Chrome for Testing
  • Telemetry: Google collects usage statistics (tool success rates, latency, environment info) by default; opt out with the --no-usage-statistics flag. Performance tools may send trace URLs to the Google CrUX API; disable with --no-performance-crux.

Security review

Review before installing

Skills for using the Chrome DevTools MCP server (debugging, performance, accessibility, cookie inspection). The server itself is not in our snapshot; it is run through npx chrome-devtools-mcp@latest, collects Google usage statistics unless you pass --no-usage-statistics, and drives a Chrome profile that can see cookies and auth headers of pages you open with it. Only the skill Markdown could be reviewed.

Worth knowing

  • Sends usage data by default (can be turned off)Google collects usage statistics (tool success, latency, environment info) by default; opt out with the --no-usage-statistics flag when starting the server. Performance tools may also send trace URLs to the Google CrUX API; disable with --no-performance-crux.README.md:45
  • Runs remote code at an unpinned versionDocumented MCP config runs npx -y chrome-devtools-mcp@latest, always executing the newest published release; skills/chrome-devtools/SKILL.md:8 also references npx chrome-devtools-mcp@latest.README.md:77
  • Relies on a CLI or package we could not reviewAll skills drive the chrome-devtools-mcp npm server and Chrome, whose source is not in the files we hold.skills/chrome-devtools/SKILL.md:8
  • Installs packages or toolsInstructions install the package globally (npm i chrome-devtools-mcp@latest -g) and require Chrome.README.md:108
  • Changes files outside your projectCLI skill install copies skills into ~/.claude/skills (cp -R "$(npm root -g)/chrome-devtools-mcp/skills/chrome-devtools-cli" ~/.claude/skills/).README.md:148
  • Makes network requestsSkills load pages, run evaluate_script (including fetch against the current URL) and read network requests, cookies and Cookie/Set-Cookie headers (including HttpOnly values) of pages in the controlled Chrome instance, which uses a persistent Chrome profile.skills/cookie-debugging/SKILL.md:55

What we checked

The repository (374 files) on Oct 29, 2026. Not a full review: 359 of 374 files were not downloaded (large repository; we reviewed the skill folders); it contains compiled binaries or archives we cannot read; it relies on code outside the repository: npm package chrome-devtools-mcp (the MCP server; 359 of 374 repo files, including all source, are not in our snapshot); Google Chrome / Chrome for Testing, and the Google CrUX API called by performance tools; tests/trace-processing fixture archives (2 binaries).

NVIDIA SkillSpector static scan: 8 pattern hits (MCP Rug Pull 3, Agent Snooping 2, Rogue Agent 2, Excessive Agency 1). SkillSpector hits are an unpinned npx mention, the README instruction to copy skills into ~/.claude/skills, a phrase 'without checking' in an LCP skill, and the memory-leak skill telling the agent to read the MCP config to see whether --memoryDebugging is on.

A review, not a certification. Skills change after we check them — read the files before installing.

License

Apache-2.0 — commercial use allowed.

Can it run unattended?

Not really. It is a tool and skill set for interactive debugging sessions; nothing in it defines a loop, stop conditions or resumable state.

Alternatives to Chrome DevTools MCP

playwright-cli

microsoft/playwright-cli

Microsoft's command-line interface for driving a browser through Playwright, plus a skill that teaches coding agents the commands.

Recommended Check first Depth 3/5

Browser Use

browser-use/browser-use

Seven skills around the Browser Use browser agent: drive local Chrome over CDP, use the paid cloud API, run QA scoring, or pay by crypto wallet.

With caveats Broad access Depth 3/5

agent-browser

vercel-labs/agent-browser

A native browser-automation CLI for agents, plus 11 skills that teach the snapshot-and-ref workflow for navigating, filling forms, scraping and testing.

Recommended Check first Depth 4/5

Web Quality Skills

addyosmani/web-quality-skills

Six skills by Addy Osmani for auditing and fixing web performance, Core Web Vitals, accessibility, SEO and best practices, built around Lighthouse.

Recommended Check first Depth 3/5

FAQ

What is Chrome DevTools MCP?

It is an MCP server and CLI from the Chrome DevTools team that lets AI coding agents control and inspect a live Chrome browser. The repo also ships seven skills for performance, LCP, memory leaks, cookies, accessibility and troubleshooting.

How do I install Chrome DevTools MCP?

Add npx -y chrome-devtools-mcp@latest as an MCP server in your client; the README links per-client guides for Claude Code, Cursor, VS Code and Antigravity. To use the CLI and skills, install the package globally with npm i chrome-devtools-mcp@latest -g and copy the skills folder into your agent's skills directory.

Does Chrome DevTools MCP collect data?

Yes, by default Google collects usage statistics such as tool success rates, latency and environment details. Start the server with --no-usage-statistics to opt out, and --no-performance-crux to stop trace URLs going to the CrUX API.

Chrome DevTools MCP vs Playwright CLI: which should I use?

Chrome DevTools MCP is built around Chrome's own DevTools, so it is strongest for performance traces, Lighthouse, memory and network debugging. Playwright CLI is a general browser-automation command set that suits scripted flows and test generation.

Sources: GitHub repository (stars, license, file tree, last push Oct 29, 2026). Review based on the README and SKILL.md files as published.