Google Antigravity opens browser control: AI can open web pages, test code on its own, and run multiple tasks in parallel.

Beating AI News Flash: Google is gradually rolling out browser control capabilities to Antigravity users. AI can now directly operate the software's built-in browser—opening web pages, clicking buttons, filling out forms, and even checking console errors and network requests.

Antigravity previously supported controlling external Chrome via `/browser`. This time, browser operations are integrated directly into the app with no additional configuration needed. For example, after AI finishes writing a website, it can open the page itself to test, discover broken buttons or layout issues, then modify the code and re-check.

The new feature also supports multiple browser Agents working in parallel. Each Agent has its own independent tab and can simultaneously research information and test different pages. Users can watch the AI's mouse operations in real time and take over at any moment.

Google has also added permission controls. The browser uses an isolated login environment and pauses when encountering logins or CAPTCHAs, letting users handle them personally. AI-initiated file downloads are also subject to approval rules—programs or scripts cannot be downloaded at will.

The feature currently supports macOS, Windows, and Linux, requires Antigravity App 2.30.0 or later, and is gradually rolling out to free, Plus, Pro, and Ultra personal accounts. Enterprise accounts and remote Agents cannot yet use the new built-in browser control.