Browser MCP serverOfficialLocal

Chrome DevTools MCP Server

Google's Chrome DevTools MCP server lets your AI agent debug a live Chrome: console, network requests, performance traces and Lighthouse audits.

By Google's Chrome DevTools team · Last updated

Set up

Terminal
claude mcp add --transport stdio chrome-devtools -- npx -y chrome-devtools-mcp@1.10.1 --no-usage-statistics

Needs Node.js. Adds it for you in this project. Add --scope user to have it in every project.

Or share it with your team in .mcp.json

.mcp.json
{
  "mcpServers": {
    "chrome-devtools": {
      "command": "npx",
      "args": [
        "-y",
        "chrome-devtools-mcp@1.10.1",
        "--no-usage-statistics"
      ]
    }
  }
}

This file is meant to be committed, so secrets are ${NAME}: Claude Code fills them in from each person's environment.

Claude Code MCP docs
Google's Chrome DevTools team's setup docsView the source

What it can do

Reads only look. Writes change something on your account, and read-only mode turns them off.

Debugging

Reads

  • List the console messages since the last page load, and read one in full
  • List the network requests, and read one's headers and response
  • Read the CSS rules that apply to an element, and which ones win
  • Take a text snapshot of the page, or a screenshot

Writes

  • Run JavaScript in the page

Performance

Reads

  • Read the detail of one insight from a trace, like what held up the largest paint

Writes

  • Record a performance trace of a page load, with Core Web Vitals
  • Run a Lighthouse audit for accessibility, SEO and best practices
  • Take a heap snapshot to look for memory leaks

Driving the browser

Reads

  • List the open tabs, and wait for text to appear

Writes

  • Open, switch between and close tabs, and go to a URL, back or forward
  • Click, hover, drag, type, press keys and fill in forms
  • Upload a file, and accept or dismiss dialogs
  • Resize the page, and emulate a device viewport, a slow network or CPU, a location or dark mode

What it can touch

Official
Yes. Published by Google's Chrome DevTools team, who make the product it connects to.
Sign-in
None. It needs no account.
Read-only mode
No. It can do anything your credentials allow.
What leaves your machine
With the flag in these snippets, no usage statistics go to Google. Performance traces may send the page URL to Google's CrUX API unless you add --no-performance-crux. What your agent reads in the browser goes to your AI tool's model.
Pinned version
chrome-devtools-mcp@1.10.1 when you run it locally. The remote server is whatever version the publisher is running.
MCP registry
io.github.ChromeDevTools/chrome-devtools-mcp
Licence
Apache-2.0
Last reviewed
How we check an MCP server is safe
Check it yourself

MCP servers checklist

0 of 11 checked

Who made it

Before reading a line of it, know whose code you are about to run.

What's inside

The part people skip. Read what your agent will read.

What it can reach

Give it the least access that still does the job.

Keeping it that way

What you checked today is only what runs tomorrow if you pin it.

Your ticks are saved in this browser only.

Chrome DevTools MCP puts the DevTools you already use into your AI agent's hands. It starts Chrome, loads your page, and reads what you would read in the Console, Network and Performance panels, so your agent can find the error or the slow request itself instead of asking you to paste it. It is made by Google's Chrome DevTools team.

Prompts to try

  • "Open localhost:3000, check the console and the failed network requests, and tell me what is broken."
  • "Check the performance of localhost:3000 and tell me what is slowing down the largest contentful paint."
  • "Run a Lighthouse audit on the signup page and fix the accessibility issues it finds."
  • "Why is this button blue? Look at the CSS rules on it."

Gotchas

  • Chrome opens only when a tool needs it. Connecting the server does not start a browser. The first prompt that needs one does, which can take a few seconds.
  • It can read anything in that browser. The README warns that it exposes the browser's contents to your AI tool. Do not sign in to accounts in its Chrome that you would not paste into the chat.
  • Two things phone home unless you stop them. Usage statistics (off in our snippets) and, during performance traces, the page URL sent to Google's CrUX API for real-user data. Add --no-performance-crux to stop the second. It also checks npm for new versions.
  • Google Chrome only. It officially supports Chrome and Chrome for Testing. Other Chromium browsers may work, without a promise.
  • Too many tools? --slim swaps in a small set for basic browsing, and the --category-* flags turn groups off.

When not to use it

If your agent only needs to click through a flow, or you test in Firefox or Safari, the Playwright MCP server is the simpler fit. For a one-off Lighthouse score, running Lighthouse yourself is quicker.

FAQ

What is the difference between Chrome DevTools MCP and Playwright MCP?

Both let your agent drive a browser. Chrome DevTools MCP is built for debugging: console messages with source-mapped stack traces, network detail, performance traces and Lighthouse. Playwright MCP is built for driving pages, works with Firefox and WebKit too, and suits testing a flow end to end.

Does it use my normal Chrome profile?

No. It starts its own Chrome with a separate profile under ~/.cache/chrome-devtools-mcp, which persists between runs. Add --isolated for a throwaway profile, or --auto-connect to attach to a Chrome you already have open, if you choose to allow that.

What does --no-usage-statistics do?

It stops the server sending Google usage statistics, such as how often tools succeed and how long they take. Collection is on by default, and our snippets turn it off. Remove the flag if you are happy to share them.

Related MCP servers

Playwright MCP Server

Microsoft's Playwright MCP server lets your AI agent drive a real browser: open pages, click, fill in forms and read each page as an accessibility tree.

See server

Filesystem MCP Server

The reference Filesystem MCP server gives your AI agent file access limited to the folders you name. Ideal for Claude Desktop and for learning MCP.

See server

Context7 MCP Server

Context7's MCP server gives your AI agent current, version-specific library docs and code examples, so it stops inventing APIs that do not exist.

See server
See all MCP servers