Skip to content
AI Tool Verdicts

Claude in Chrome Extension: Setup, Plans and Safety

Claude in Chrome explained for October 2026: what the extension can do, which plans get it, how to install it, its approval modes and how to use it safely.

By AI Tool Verdicts Editorial Team · Published

A laptop on a desk showing a plain browser layout with a terracotta side panel, next to a brass compass, illustrating Claude in Chrome

Bottom line: Useful for repetitive web chores and for developers testing their own apps. Run it in a separate Chrome profile and keep it away from banking and sensitive accounts.

Claude in Chrome is Anthropic's browser extension that lets Claude use Chrome for you. It reads the page you're on, clicks, types, fills in forms and works across several tabs, all in a side panel. As of October 2026 it's available on every paid Claude plan (Pro, Max, Team and Enterprise), but not on Free, and officially only in Google Chrome on desktop.

It's good at repetitive web chores: pulling data from pages, filling forms, sorting through a busy inbox. It's also a real security trade-off, because anything Claude reads on a web page could try to hijack it. Below are the setup steps, the three approval modes, and the safe defaults we'd use.

How we researched this: Anthropic's Claude in Chrome page, its help center articles on setup, safety and permissions, the Claude Code Chrome docs and Anthropic's launch posts, checked October 28, 2026. We haven't run our own tests of the extension.

What can the Claude Chrome extension do?

Claude works inside the tabs you give it, using the sites you're already signed in to. Per Anthropic's getting started guide, it can:

  • Read pages and act on them: open links, click, type and fill forms.
  • Work across multiple tabs, which it keeps in its own Chrome tab group.
  • Record a workflow once and replay it as a saved shortcut, called with "/".
  • Run scheduled tasks, such as a daily summary of new messages.
  • Keep working in the background while Chrome is open.
  • Take screenshots and use images you share with it.
  • Find its way around Slack, Gmail, Google Calendar, Google Docs and GitHub with built-in knowledge of those sites.

It also connects to the Claude desktop app, so a chat or task there can use your browser. Tasks that touch files on your computer need the desktop app open and connected.

Who can use Claude in Chrome?

Details (October 2026)
PlansPro, Max, Team and Enterprise. Not Free.
BrowserGoogle Chrome on desktop. Anthropic says it's not supported on other Chromium browsers or on mobile.
Team and EnterpriseAdmins can turn it on or off and set allowed and blocked sites.
Healthcare (HIPAA) Enterprise orgsOff by default; an owner must enable it and allow specific sites.
Via Bedrock, Vertex or FoundryNot available.

There's one wrinkle on browsers. The extension's own help page says Chrome only, but the Claude Code docs say the Claude Code integration works with Chrome and Edge, and can detect the extension in Brave, Arc, Vivaldi and Opera. For everyday use, plan on Chrome.

The rollout took about four months, per Anthropic's launch post:

DateWhat changed
August 25, 2025Research preview for 1,000 Max users
November 24, 2025Beta for all Max subscribers, with scheduled tasks and multi-tab workflows
December 18, 2025Expanded to Pro, Team and Enterprise, plus Claude Code integration

How to install Claude in Chrome

  1. Open Google Chrome on your computer.
  2. Find Claude in the Chrome Web Store and click Add to Chrome.
  3. Sign in with your paid Claude account.
  4. Pin the extension: click the puzzle piece icon in the toolbar, then the pin next to Claude.
  5. Review and grant the permissions it asks for.

The extension requests 15 Chrome permissions. The ones worth understanding are debugger, which lets it click, type and take screenshots, and scripting, which lets it read page text. Those are what make it work, and also why it needs care.

Before step 1, we'd add one step of our own: create a separate Chrome profile for Claude. More on why below.

Claude in Chrome approval modes

The extension has three modes, chosen from a dropdown in the chat box. The names come from Anthropic's permissions guide.

ModeWhat happensUse it for
Manually approveClaude pauses and asks before each actionAny site you haven't used it on before, anything involving accounts
Automatically approveA safety check reviews each action, blocks unsafe ones and pauses when needed. The default.Routine tasks on trusted sites
Skip all approvalsClaude doesn't pause, and nothing checks its actions automaticallyThrowaway tasks on low-risk public sites only

Each site also has its own permission. When Claude reaches a new site, you can allow one action, always allow actions on that site, or decline. Even on an always-allowed site, it still asks before downloading a file, entering sensitive information or granting an authorization.

You can review and revoke approved sites in the extension's settings, along with your permission history.

What won't Claude in Chrome do?

Anthropic's permissions guide lists actions the extension won't take in any mode:

  • Make purchases or financial transactions, or execute trades.
  • Create accounts, or enter credit card or ID numbers.
  • Permanently delete things, such as emptying the trash or deleting emails.
  • Download files from untrusted sources or modify system files.
  • Follow instructions it finds inside emails or web pages.
  • Bypass CAPTCHAs.

It also blocks adult and pirated content sites, and asks first on financial sites. When it reaches a login page or a CAPTCHA, it stops and hands control back to you.

Is Claude in Chrome safe?

Safer than it was, but not risk-free. Anthropic says so itself on its safety page: "the risk is not zero."

The main threat is prompt injection: text hidden in a web page or email that tells Claude to do something you didn't ask for. Anthropic's numbers show how far its defenses have come:

  • In August 2025, with no defenses, attacks in Anthropic's tests succeeded 23.6% of the time. Its first safeguards cut that to 11.2% in autonomous mode.
  • As of October 2026, Anthropic's safety page says two classifiers now screen incoming content and check each action before it runs, and reports an attack success rate under 0.08% in its internal testing.

Those figures come from different models and test sets, so they're not a clean trend line. But the direction is clear, and the remaining risk is why the setup matters.

The safe setup we'd recommend, based on Anthropic's own guidance:

  1. Use a separate Chrome profile with no saved passwords and no access to email, banking or work admin accounts.
  2. Stay in Manually approve until you've watched Claude handle a task a few times.
  3. Grant site access one site at a time, not "always allow" by default.
  4. Stop the task if Claude visits a site you didn't expect or asks for sensitive information.
  5. Don't use it for banking, health records, legal matters or other people's personal data.
  6. On Team or Enterprise, ask your admin for an allowlist of approved sites.

Claude in Chrome with Claude Code

For developers, this is the most useful part. Run claude --chrome and Claude Code can open your app in the browser, read console errors and the DOM, test forms, upload files and record GIFs of what it did. It's a fast way to let the agent check its own front-end work.

A few details from the Claude Code docs:

  • You need extension version 1.0.36 or later, and Claude Code logged in with /login. API-key logins don't get the Chrome integration.
  • The extension and Claude Code must be signed in to the same Claude organization.
  • Run /chrome inside a session to check the connection, reconnect or turn it on by default. Turning it on by default uses more context.
  • File uploads are capped at 10 MB, and it refuses files like .env and .pem.

If you're choosing a coding agent, our Cursor vs Claude Code comparison and Claude Code pricing guide cover the rest of the setup.

Common problems and fixes

  • The extension stops responding in a long session. Chrome's service worker can go idle. In Claude Code, run /chrome and choose Reconnect extension.
  • A pop-up dialog blocks everything. JavaScript alerts and confirm boxes stop the browser tools; close them yourself.
  • Your shortcuts and scheduled tasks disappeared. Logging out of the extension removes them.
  • It works on your laptop but not at work. The extension needs to reach bridge.claudeusercontent.com, which some company firewalls block.

Is Claude in Chrome worth it?

Yes, if you're already on a paid plan and have web tasks you repeat every week. It costs nothing extra, and saved shortcuts turn a ten-minute routine into one command. It's also excellent for testing your own web app with Claude Code.

Skip it if your work happens mostly in banking, healthcare or admin consoles, where one wrong click is costly. If you're on Free and deciding whether to upgrade for this, our Claude pricing guide shows what each plan adds.

FAQ

Is Claude in Chrome free?

No. Claude in Chrome needs a paid Claude plan: Pro, Max, Team or Enterprise. The extension itself costs nothing extra on those plans.

Does Claude in Chrome work on Edge, Brave or Safari?

Officially, the extension supports Google Chrome only. Claude Code's browser integration also works with Edge and detects the extension in some other Chromium browsers. Safari and mobile browsers aren't supported.

Can Claude in Chrome make purchases or enter passwords?

No. It won't make purchases, execute trades, create accounts or enter card details. It pauses at login pages for you to sign in yourself.

How do I turn off Claude in Chrome?

Remove or disable the extension from Chrome's extensions page. To stop it on one site, revoke that site in the extension's settings under approved sites.