What this does. Claude in Chrome is a browser extension that lets Claude see a web page and act on it: open tabs, click buttons, type into fields, read what is on screen, and chain those actions into a workflow. It uses the logins you already have in your browser, so there are no API keys to wire up.
The honest origin story. The classroom you are reading this in was built with this exact tool. I did not click through Skool by hand. I told Claude what each course and lesson should say, and it opened Skool, created the courses, filled the lesson editor, and saved each page. This lesson shows you how to set up the same thing.
Requirements: a paid Claude plan (Pro, Max, Team, or Enterprise) and Google Chrome. Browser control is a beta feature.
Open Chrome and go to the Chrome Web Store listing for the extension (it is named Claude):
The standalone extension runs in Google Chrome. The Claude Code integration (next step) also works in Microsoft Edge. It does not work on Brave, Arc, other Chromium browsers, mobile, or WSL.
Claude only acts on sites you allow. The extension asks for permission per site, and you manage the list in the extension's settings. Approve a site the first time you want Claude to work on it, and leave everything else off.
When Claude hits a login screen or a CAPTCHA, it stops and hands control back to you to finish by hand. It will not enter passwords or solve bot checks for you. That is the safety model working as intended.
There are three ways to drive the browser. Pick whichever fits.
Open the Claude side panel in Chrome and talk to it about the page you are on. Best for quick, in-the-moment tasks while you browse. Turn on Ask before acting mode to have Claude write a plan, get your approval, then run the whole workflow on its own.
If you use Claude Code, you can give it browser control so it can build code and then test it in a real browser. Start a session with the flag:
Or turn it on inside a running session:
This needs the Claude in Chrome extension version 1.0.36 or higher and Claude Code version 2.0.73 or higher. In the VS Code extension, Chrome is available automatically once the extension is installed, with no flag.
The Claude Desktop app can use the extension as a connector, so it does browser work without you switching windows. This works in a normal Chat, where you guide it step by step, and in Cowork, where you describe the outcome and let it run. Set it up once:
Then just ask. A simple first test, in a Chat or a Cowork task:
Paired with Cowork, Chrome becomes the research layer: Claude gathers what it needs from the web, then Cowork turns it into finished files like spreadsheets and decks, with the context flowing between them automatically.
Start read-only so you can watch how it behaves before letting it act. In the Claude Code CLI (after /chrome) or the side panel, try:
Claude opens a visible Chrome window and you watch it work in real time. When you are comfortable, move on to real tasks: filling a form from a spreadsheet, pulling data off a page into a CSV, or drafting into a Google Doc you are signed into.
Extract a list off a web page into a CSV. Fill a repetitive form from a spreadsheet. Draft a reply in Gmail or a doc in Google Docs. QA your own web form by submitting bad data and checking the error messages. Record a short GIF of a workflow to share.
Run /chrome any time to check status, reconnect, manage permissions, or pick which browser to use.
--chrome every session, run /chrome and choose Enabled by default. (This loads browser tools every session and uses more context, so turn it back off if you notice that.)/chrome and choose Reconnect extension."Extension not detected"
Confirm the extension is installed and enabled at chrome://extensions, that Claude Code is current (claude --version), and that Chrome is running. The first time you enable the integration, Claude Code writes a configuration file that Chrome reads on startup, so restart Chrome if it is not detected on the first try.
Browser stops responding mid-task
A JavaScript dialog (alert, confirm, or prompt) blocks the page and freezes browser control. Dismiss the dialog by hand, then tell Claude to continue.
Connection drops on a long session
Run /chrome and choose Reconnect extension.
Which model runs?
On Pro, browser tasks use Claude Haiku 4.5. On Max, Team, or Enterprise, you can choose Opus 4.7, Sonnet 4.6, or Haiku 4.5.
I only access Claude through Bedrock, Vertex, or Foundry
Browser control is not available through third-party providers. You need a direct claude.ai account (Pro, Max, Team, or Enterprise) to use it.
Made by AI Service Engine · official Chrome docs · no terminal? try Cowork next