Gemini can now take over your Mac to run multi-step tasks
Google rolls out computer use in the Gemini app for Mac, running in an isolated background window so it never touches your cursor while it works.
Google started rolling out computer use in the Gemini app for Mac on 19 November. It lets Gemini complete multi-step tasks across your local files, desktop apps and the web: navigating menus, entering text and filling in forms on your behalf.
The agent runs inside an isolated background window with a live preview in the chat, so it doesn’t take over your main screen or cursor while you keep working. You choose which apps it can access, watch each step as it happens, and approve key actions before they run. It needs an Apple Silicon Mac on macOS Sequoia 15.0 or newer, a personal Google account for someone 18 or over, and a Google AI Plus, Pro or Ultra subscription. For now it’s English only, in the US.
Why it matters: this is the same computer-use pattern Anthropic and OpenAI have been shipping through their APIs, but aimed at Google’s own consumer app rather than developers building agents. If you’re building or evaluating computer-use agents, it’s worth a look at how Google handles the permission model: per-app access, a visible live preview, and approval gates before action, rather than silent background control.
The caveat: it’s a US-only, English-only rollout gated behind a paid subscription tier, so most people reading this won’t be able to try it yet.