Claude Code Gets a Built-In Browser on Desktop

Anthropic has added a built-in web browser to the Claude Code desktop app. The coding agent can now open documentation pages, design files, or bug reports on its own, read their contents, click elements, and fill in fields, without having to switch to a separate Chrome window.

Until now, Claude Code could only display a preview of a locally running application, essentially what a developer sees while working on their own code. The new feature extends that ability to the entire web: the agent can pull up a library's documentation, check how a design looks in Figma, or review a bug report on GitHub, then return to the code it's editing and make changes.

How It Works in Practice

The browser launches as a tabbed panel inside Claude Code, alongside the preview of the local development server. The agent operates it the same way it has always operated previews of its own applications: reading page content, clicking buttons, filling out forms, and reading back the results of its actions.

Claude can open documentation, design files, or any other webpage. It can read it, click on it, and interact with it just as it does with local development servers - Anthropic, announcement on the ClaudeDevs account

The panel can also handle login pop-ups, including Google OAuth, but it does so on a clean browser profile with no saved passwords or user history. In other words, the agent logs in fresh rather than impersonating the computer owner's account. Anyone who needs Claude to operate within their own logged-in browser session still has to use the separate Chrome extension that Anthropic has offered for some time.

Security and Organizational Controls

Because the agent gains the ability to navigate freely across any website, Anthropic added extra layers of control. Safety classifiers analyze every write action carried out on external sites, and Claude cannot make a purchase, create an account, or bypass a CAPTCHA without the user's explicit consent.

Companies using Claude Code across teams can restrict the browser to an allowlist of domains or disable the tool entirely for their employees. That matters for security teams, which have already had to respond to other Claude Code features that raised questions about how much access coding agents should be granted.

The Race for Coding Agents

The built-in browser arrives as Anthropic's rivals test similar approaches to connecting agents with the web. A week earlier, OpenAI shut down its standalone ChatGPT Atlas browser, folding page-browsing features into its integrated desktop app. The competition now centers on which agent can more effectively combine code writing with real testing of interfaces and documentation without leaving a single window.

For developers, this means less app-switching during everyday work: the agent can check what the API documentation looks like, test a form on a staging site, or verify a bug report before proposing a code change.

What It Means for Polish Teams

Claude Code is widely used by Polish development teams, including in large enterprise deployments, where both the agent's convenience and control over what resources it can access matter. The ability for organization administrators to block the browser addresses concerns from security teams, who have to balance team productivity against the risk of accidental data exposure or an unwanted action on an external site.

The feature is now available to all Pro and Max plan subscribers after updating the app to version 2.1.202 or later, at no extra cost beyond the standard subscription.

Sources: The Decoder (the-decoder.com), 9to5Mac (9to5mac.com), MLQ News (mlq.ai)