How does the Antigravity browser subagent work?

What it is

A browser subagent is a helper agent that operates a real browser for the main agent. In Antigravity, you call it with /browser.

How it works

1. In the app or CLI, run /browser with a task, such as “sign up with a test account and report any errors”.
2. The subagent drives Chrome through the Chrome DevTools MCP server: it navigates, clicks, types and reads the page.
3. It reports results back to the main agent, which can fix code and re-test.
4. In the Antigravity IDE, use the IDE’s own browser features instead.

Setup (checked 2026-10-16). Google says no manual setup is needed: Antigravity starts the Chrome DevTools MCP server itself when the subagent is called. The first time it connects, Chrome asks you to allow remote debugging; accept so the subagent can attach. Each session is recorded and saved to the conversation’s artifacts as a WebM video. Your organization’s admin can disable browser access, in which case /browser doesn’t appear.

Benefits

• Tests real user flows, not just unit tests.
• Lets the agent verify its own UI changes before you review.
• Useful for checking forms, signups and checkout flows on marketing sites.

Limits and risks

• Use test accounts; don’t let a subagent act in production accounts without approval.
• Sites may block automation or show CAPTCHAs.
• Not available in the IDE surface.