Google Adds Computer Use to Gemini for Mac

Google has added computer use to its Gemini desktop app for macOS, letting the assistant navigate native Mac apps and browse websites in the background to complete tasks for a user. TechCrunch reported the launch on November 19, 2026.
The feature targets multi-step work that spans local files, applications, and the web. Google lists three example uses: filling out forms, sorting files into folders according to specified criteria, and creating slideshows from files or from information in online dashboards. Google's Gemini blog carries a post titled "Let Gemini tackle repetitive chores on your Mac," written by Michael Friedman, Group Product Manager, Gemini App. The post frames the feature around recurring, low-judgment tasks.
Gemini displays its computer-use activity as a live preview inside the chat window. The user's screen and cursor stay free for other work while the agent operates. Users can pause or stop a run at any point, and they can approve computer-use actions before they execute.
Computer use is initially available in English in the United States. It requires the Gemini app for Mac and a Google AI Plus, Pro, or Ultra subscription.
The broader context here is the shift in where agentic behavior runs, meaning software that carries out a sequence of actions toward a goal on its own. A chat assistant that calls APIs, the structured interfaces services expose to other software, operates inside the boundaries of the services it connects to. An agent that drives the desktop, clicking, typing, and navigating the same interface a person uses, works with whatever the signed-in account can reach. That makes the control surface central. The pause, stop, and pre-execution approval steps are the mechanisms Google has described for keeping a person in the loop, and they are the parts of the design that enterprise security teams and careful individual users will examine first.
In my view, the background operation matters as much as the capability itself. Most automation tools either demand the user's attention for the duration of a task or run unattended in a batch. Showing progress in the chat while leaving the screen and cursor available sits between those two modes. That is why a task like sorting a downloads folder or assembling a deck from a dashboard can run while the user writes something else. Whether that balance holds across real desktops, with unfamiliar apps, inconsistent layouts, and sites that change without notice, is a question the early rollout will answer.
The feature also sets a clear boundary on who can use it today. Access depends on a paid Google AI tier, a Mac, and the English-language U.S. market. Those limits are the first thing an IT administrator or a power user will notice, and they determine how quickly computer use spreads across a workforce.
For the Mac user, the practical change is narrow but concrete. Tasks made up of many small, rule-driven steps across several windows, the kind people tend to postpone, can now be delegated to an assistant that reports its progress and waits for approval before acting. If the controls work as described, that is a meaningful reduction in the friction of routine desktop work, and it is the kind of everyday gain that tends to matter more over time than any single demonstration.


