Claude Code gets a built-in sandboxed browser inside the desktop app

Claude Code now has its own browser.

Anthropic added a built-in, sandboxed browser to the Claude desktop app, as reported by 9to5Mac on July 10, 2026. Claude Code can pull up docs, designs or any other site, read pages, click through, and interact with them the same way it already works with your local dev servers.

Sandboxed by design

The privacy architecture is the notable engineering choice. The built-in browser runs a clean profile, separate from your personal browser: none of your saved logins, none of your history. Sessions are sandboxed and configurable, and you choose whether they persist.

For browsing that needs your accounts, Anthropic points to the Claude Chrome extension instead. The split is deliberate: the agent’s browser is a clean room for testing and reading, while logged-in browsing stays in your own browser under your own control.

Closing the loop

The reason this matters for developers is what it completes. An AI coding agent that can write code, run it, open the result in a browser, see what broke and fix it is a full build-test-debug cycle inside one app.

Until now, that last step, actually looking at what the code produced, was where the human had to step back in. With a browser in the loop, Claude Code can verify its own work the way a developer does: load the page, click the button, watch it fail, and go fix it.

Pointed timing

The competitive context is hard to miss. The same week OpenAI shut down its standalone Atlas browser, Anthropic put a browser inside Claude.

Two rival bets on the same question resolved in the same week: OpenAI concluded a browser is not a product worth maintaining on its own, while Anthropic concluded a browser is exactly the tool its agent was missing. Notably, both agree on where browsing belongs, inside the AI app rather than around it. Would you let an agent browse and test for you? For Claude Code users, that option now ships in the box.