PlugWiki
Browser MCP logo

Browser MCP Server: Let Claude and Cursor Use Your Chrome

CommunityLocal
Getnpm@browsermcp/mcpbrowsermcp/mcp
Tools
1
0 read · 1 change
Works in
12
apps
Risk
High
can delete
Auth
None

Lets Claude, Cursor and VS Code use your own Chrome, in your signed-in tabs, through the Browser MCP extension.

Browser MCP lets an AI app work in your real Chrome, including sites where you are already signed in. It has one tool that takes browser commands, so the assistant can open pages, click, fill forms and read what is on screen. Everything runs on your computer, but you first install the Browser MCP extension in Chrome and click Allow once after checking a six-digit code. It acts as you on every site your profile is signed in to, so treat it as full access to those accounts.

Good for

  • Developers who want an assistant to test or script pages in a real, signed-in browser
  • Anyone who wants Claude to do routine tasks on sites that need a login
  • Researchers who need pages read from their own browser session
  • Cursor and VS Code users who already run Node.js

Works in

12 should work

Tap an app to see how to set it up and any known problems.

Install Browser MCP

Claude Code

  1. 1

    Copy this prompt.

    Add the Browser MCP server to Claude Code. Run this command:

claude mcp add browsermcp -- npx -y @browsermcp/mcp@latest

Then check that it loaded and tell me if I need to sign in or restart.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

Install the Browser MCP extension in Chrome first (from the Chrome Web Store, linked on browsermcp.io) and pin it. The first time the assistant uses the browser, check that the six-digit code in the extension popup matches the one it shows, then click Allow. The package name ends in @latest, which the project's README uses so you always get the newest server.

Built from the package data we hold. Not yet tested by us.

Tools

  • Runs browser commands in your own Chrome, one per line

    Controls your Chrome through the extension. The assistant sends one command per line: go to a page, click, type, press a key, take a screenshot, read text or run a JavaScript function in the page. It stops at the first error. Uploads and screenshots can read and write files on your computer, and a raw browser-protocol mode stays off unless you turn it on.

    Try: “Open my company's staging site, log in with the saved session and tell me what the dashboard shows.”

    Runs code · browser

Read from the source code

Permissions

high risk · 4 things to watch

High risk

The assistant can click, type and run scripts in your signed-in Chrome, so it can do anything you can do on those sites, including posting or deleting. Use a separate Chrome profile for it, keep it away from banking and admin accounts, and watch what it does.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • It acts as you on signed-in sites

    Tabs the assistant opens use your saved logins, so it can send, buy or delete wherever you could. Use a separate Chrome profile with only the accounts it needs. github.com

  • Pages can steer it to files on your computer

    The README warns that upload and screenshot paths are not limited, so a page the assistant reads could make it upload any file you can read or overwrite any file you can write. Browse only pages you trust. github.com

  • It runs JavaScript in pages

    An eval command runs a JavaScript function inside the page it is on. That is useful for testing and risky on untrusted sites. github.com

  • Incognito and browser pages stay blocked

    The extension refuses incognito tabs unless you allow them, browser and extension pages, and everything after you press Stop. Use Stop if the assistant goes off track. github.com

Security advisories

No published security advisories as of Nov 1, 2026.

Technical details
Turning on read-onlyNo read-only mode is documented. The tool can click, type, submit forms and run JavaScript on any page the assistant opens.
Talks toAny site you open. Local daemon on 127.0.0.1
Sign-in and keysNo account or API key. You install the Chrome extension and click Allow once per computer after checking that its six-digit code matches the one your AI app shows.
Scopes

  • Your Chrome profile and every site it is signed in to

Questions

Do I need the Chrome extension?

Yes. Install the Browser MCP extension from the Chrome Web Store, pin it, then add the server to your AI app. The first time, click Allow in the extension popup after checking that its six-digit code matches the one your AI app shows.

Which browsers does it work in?

Chrome and other Chromium browsers such as Edge, Brave and Arc. Firefox is not supported because it has no debugger API.

Does it use my logged-in sessions?

Yes. It works in your own Chrome profile, in tabs of their own, so sites where you are signed in stay signed in.

Is it safe?

It runs on your machine and talks only to a local daemon, but it can act as you on any signed-in site. Use a separate profile and watch what it does.

Does it need an API key or an account?

No. It needs Node.js 20 or later for the server and the Chrome extension, and no account.

Can I build the extension myself?

Yes. The README explains how to build it from the repository and load it unpacked, which needs Node.js 24 and pnpm.

Is it official?

It is the project's own server, made by the Browser MCP project. It is not made by Google.

Information

updated Nov 2, 2026

Publisher
Community
Package
@browsermcp/mcp
Registry
npm
Transport
stdio
Auth
None
License
Apache-2.0
Last commit
Oct 30, 2026
Open issues
0

WebsiteGitHubIssues

Sources

Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.

Show all 2
  1. github.com · github.com
  2. github.com · github.com

Facts from the official MCP registry, GitHub, npm, PyPI and the Docker catalog. Updated Nov 2, 2026.