Anthropic’s Claude now has a browser of its own
The New Stack Frederic Lardinois
Claude now has its own built-in browser on desktop. Anthropic wants web tasks inside the app, not poking through your main browser.
Based on reporting by The New Stack, Frederic Lardinois — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic has put a browser inside Claude’s desktop app. On Mac, Windows, and Linux, paying Pro, Max, and Team users are getting a built-in Chromium-based browser in Cowork, so Claude can work the web without leaning on a separate Chrome extension first.
That puts Anthropic closer to OpenAI’s approach, after OpenAI backed away from building Atlas as a stand-alone browser and folded browser access into ChatGPT desktop instead. For people who’ve used the ChatGPT app with a built-in browser, the appeal is obvious: you can jump into a page, grab a screenshot, and keep talking to the model without bouncing between tools.
Anthropic still isn’t throwing away the old path. If Claude needs to work on a page that’s already open in your browser, the Chrome extension remains the better fit, especially for things like inbox work, CRM updates, or editing a document you already have in front of you. And if the extension is installed, it stays the default unless you change the preferred browser in Claude Desktop settings.
The bigger shift is that Claude now has “its own browser,” but not your browser. That matters because the new setup starts without your day-to-day logins. Anthropic says you can bring logins over from Chrome, Edge, and Firefox on macOS, and from Firefox on Windows and Linux, but banking, email, and single sign-on accounts are left out by default.
Anthropic is blunt about the tradeoff: browser agents can still get tricked by prompt injections, even with safeguards in place, and “the risk is not zero.” Still, if the browser isn’t carrying your most sensitive accounts, the damage from a bad run should be a lot smaller than if Claude were loose in your main browser profile.
My take — AI-written commentary, not fact-checked reporting
This is the right kind of boring product move: fewer hacks, fewer weird extensions, more direct control. The real story is that every serious AI assistant is getting dragged toward the browser, because that’s where the work and the mess already live. And yes, “the risk is not zero” is the sort of line that should make everyone read the settings twice.
Read more about this at: The New Stack