Google launched a computer-use feature Thursday that lets its AI assistant open applications, navigate menus, fill out forms, and move information between local files and websites on a user’s behalf.
The feature runs in an isolated background window on Mac, leaving the main display and cursor available for the user. Users can watch agent actions through live preview in the Gemini chat interface and interrupt tasks at any time.
“Starting today, computer use in the Gemini app for Mac helps you get multi-step tasks done across your local files, desktop apps, and the web,” Google said in a company announcement. “Just describe what you need, and Gemini handles the clicking, typing, menus, and forms for you.”
Access controls and permissions
The feature requires an Apple Silicon Mac running macOS Sequoia 15.0 or later. Users enable computer use in the Mac app, open Task mode, and prompt Gemini with @Computer use.
Before controlling a new application, the assistant asks permission with three options: deny, allow once, or allow ongoing. Users can create allowed and blocked app lists, monitor the virtual screen, and stop tasks. Sensitive actions require confirmation before the agent proceeds.
The rollout is limited to English in the United States for adult users with personal paid Google accounts. Work and school accounts are not supported at launch.
Competitive landscape
Competitors have developed similar tools. OpenAI offers Codex, Anthropic has built Claude with computer-use capabilities, and Kimi Work provides a comparable product. Google previously introduced Gemini Spark, a cloud-based task agent.
Timing and security context
The launch comes three days after Apple introduced tighter access controls in macOS 27.3 beta. File backups created by the feature remain available for 24 hours.
Users can monitor agent activity in real time and retain full control to halt operations. The isolated background window architecture preserves the user’s active workspace while Gemini executes delegated tasks.