Google is adding computer-use capabilities to the Gemini desktop application on macOS. The agent can navigate native applications, local files and websites to complete multistep tasks while the user continues working with the main screen and cursor.
A live preview inside the Gemini chat shows what the agent is doing. Users can pause or stop the run and approve actions before they occur. Google suggests workflows such as organizing batches of files, completing forms, repeating a task across documents or building a presentation from local and online information.
The release brings desktop control together with Gemini Tasks, which can already sign into websites, make purchases and book appointments. That broader access increases the consequences of a bad interpretation or malicious content encountered on a page. Approval prompts help only when they clearly describe the exact action, target and data involved.
Computer use is initially available in English in the United States through the Gemini app for Mac. It requires a Google AI Plus, Pro or Ultra subscription, and Google has not announced timing for other markets or operating systems. Users should begin with reversible, low-risk work and avoid granting broad access to private messages or financial accounts until the permission and audit model has been tested in practice.