Free guide
Chrome DevTools MCP: Let Your Coding Agent See the Page It Builds
Google's free, open-source way to give your coding agent a real Chrome window and the DevTools you already use. What it does, how it works, how to set it up, and the fine print.

The short version
- Google's Chrome team says coding agents are "effectively programming with a blindfold on": they never see what their code does in the browser. This tool lets them look.
- Your agent gets its own Chrome: it can click and fill forms, read console messages, check network requests, take screenshots and record performance traces.
- It's trending on GitHub this week (+3,306 stars, 58,588 total on Nov 21). Version 1.13.0 (Nov 11) added logpoints and network mocks.
How it actually works
Taking the blindfold off
Without it, your agent edits code and tells you it's fixed, and you are the one who opens the browser, finds the error and pastes it back. With it, the agent opens the page itself and checks its own work, using the same DevTools you'd use.
An MCP server with DevTools inside
MCP is the open standard agents use to call outside tools. This server gives your agent tools like navigate_page, click, fill_form, list_console_messages, list_network_requests, take_screenshot and performance_start_trace. Under the hood it drives Chrome with Puppeteer and reads traces with Chrome DevTools.
What Google's demo shows
In Google's launch demo, Gemini CLI starts Chrome, records a performance trace of a local page, names the bottleneck for the Largest Contentful Paint as resource load delay caused by a low-priority image, and edits that one image line, setting its fetchPriority to high.
Logpoints: sticky notes on a line of code
New in 1.13.0: set_logpoints lets the agent pin an expression to a line of your code. Every time that line runs, the value is logged to the console, without pausing the page or changing the source. The agent then reads the output with list_console_messages.
How to set it up
1.Check you have Node.js, npm and Chrome
It needs a Node.js LTS version, npm and the current stable Chrome (or newer). Chrome and Chrome for Testing are the officially supported browsers.
node --version npm --version2.Add it to your agent
Pick the line for your agent. For Cursor and other MCP clients, add the JSON config in their MCP settings instead.
# 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 # Gemini CLI (MCP + skills) gemini extensions install --auto-update https://github.com/ChromeDevTools/chrome-devtools-mcp # Any other MCP client (JSON config) {"mcpServers":{"chrome-devtools":{"command":"npx","args":["-y","chrome-devtools-mcp@latest"]}}}3.Or install it as a Claude Code plugin with its skills
The plugin adds the MCP server plus skills like debug-optimize-lcp, a11y-debugging and memory-leak-debugging. If you already added the MCP server with the command above, remove it first. Restart Claude Code afterwards and check with /skills.
/plugin marketplace add ChromeDevTools/chrome-devtools-mcp /plugin install chrome-devtools-mcp@chrome-devtools-plugins4.Give it its first job
This is the README's test prompt. Your agent should open Chrome and record a performance trace. Chrome only starts when the agent first uses a browser tool.
Check the performance of https://developers.chrome.com5.Then point it at your own site
Prompts from Google's launch post you can copy. Swap in your own local address.
Verify in the browser that your change works as expected. A few images on localhost:8080 are not loading. What's happening? Localhost:8080 is loading slowly. Make it load faster.
The fine print
- Free and open source under the Apache-2.0 license. It runs on your machine through npx; your agent and its model are billed as usual.
- The README warns it exposes the browser's content to your agent, which can inspect, debug and modify any data in it. Don't sign that browser into accounts you wouldn't hand to your agent. It uses its own Chrome profile by default (in ~/.cache/chrome-devtools-mcp), and --isolated gives a temporary one that is wiped when the browser closes.
- Google collects usage statistics by default (tool success rates, latency, environment). Turn it off with --no-usage-statistics.
- Performance tools may send trace URLs to Google's CrUX API to add real-user data. Turn it off with --no-performance-crux.
- Some tools need a flag: network mocks need --categoryNetworkMocks=true, and the memory heap-snapshot tools need --memoryDebugging=true.
FAQ
Is Chrome DevTools MCP free?
Yes. It's open source under the Apache-2.0 license and runs locally with npx. You still pay for the agent or model you use with it.
Does Chrome DevTools MCP work with Claude Code?
Yes. Run claude mcp add chrome-devtools --scope user npx chrome-devtools-mcp@latest, or install it as a plugin with /plugin marketplace add ChromeDevTools/chrome-devtools-mcp and /plugin install chrome-devtools-mcp@chrome-devtools-plugins.
Which agents does it support?
Any MCP client. The docs have setup steps for Claude Code, Codex, Cursor, Gemini CLI, Copilot / VS Code, Antigravity and more.
Is it safe to use with my normal Chrome logins?
Be careful. Your agent can read and change anything in the browser it controls. By default it starts its own Chrome profile, not your everyday one, so keep personal accounts out of it, or use --isolated for a throwaway profile.
What are logpoints in Chrome DevTools MCP?
Added in version 1.13.0 (Nov 11, 2026), a logpoint logs an expression to the console every time a line of code runs, without pausing or changing the source. The agent sets them with set_logpoints and reads them with list_console_messages.
Sources
- Chrome DevTools MCP on GitHub
- Setup for each agent (client configurations)
- Tool reference
- Google's launch post: Chrome DevTools (MCP) for your AI agent
- v1.13.0 release notes
Facts checked 2026-11-21 against the chrome-devtools-mcp README, LICENSE, tool reference, client configuration and configuration docs, the v1.13.0 release notes, Google's Sept 23, 2025 launch post and demo video, and the GitHub API (58,588 stars) plus GitHub Trending weekly (+3,306).
Want help building with tools like this?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolMore free guides
GPT-6.2 Drops Its Reasoning When You Edit Old Messages: How to Keep It
Every reasoning item is now tied to the conversation before it. Edit that history and the API drops the reasoning that follows. Here's how it works and how to keep it.
Inkling 2 Mini: Run Thinking Machines' Looped AI Model on Your Mac
Thinking Machines' new open-weights model runs on a 32GB Mac with the wifi off. What it is, how the loop trick works, how to set it up, and the fine print.
fast-check: Property-Based Testing for JavaScript and TypeScript, Explained
A free, MIT-licensed testing library that tests a rule instead of a few examples. What it does, how it works, how we caught a one-cent bug we planted in a test app, and how to set it up on your own code.