Google brings Gemini computer use to the Mac
Google’s Gemini app can now navigate Mac apps, local files and websites for supervised multi-step workflows.
Gemini can now carry out multi-step tasks on a Mac, with Google’s app navigating desktop software, local files and websites while the user watches a live preview. The feature is rolling out in English in the United States for eligible subscribers, and Google says it keeps the user in control of permissions and approvals.
Google announced the feature on 19 November through the Google Blog. It is part of the Gemini app for Mac rather than a new standalone operating system tool. The company says users can describe a job in ordinary language, then let Gemini handle the clicking, typing, menu navigation and form entry needed to complete it.
What Gemini computer use does
The feature is aimed at repetitive workflows that usually involve moving information between several apps or services. Google’s examples include colour-grading batches of photos, setting up complex 3D scenes, extracting measurements and addresses from local PDFs to fill in shipping forms, and collecting dashboard data for a presentation in Google Slides.
The Google demonstration focuses on a child’s butterfly sketch. Gemini is asked to turn the sketch into a model and rig it in Blender, giving the announcement a concrete example of the sort of multi-step desktop job the company has in mind. The app can work across local files, installed applications and the web, rather than stopping at a single chat response or browser page.
Google says Gemini operates in an isolated background window. That allows the user’s main screen and cursor to remain available for other work while the agent operates. A live preview appears in the chat so the user can follow what is happening instead of handing over an invisible process.
Permissions and approvals
The controls are important because computer-use tools can do more than retrieve information. Google says users choose which applications Gemini can access and can watch the agent’s actions as they happen. The user can pause the process, while key actions require approval before they are carried out.
That model makes the feature closer to a supervised assistant than a completely autonomous desktop replacement. It can perform the mechanical part of a workflow, but the user still decides which apps are available and whether sensitive steps should go ahead. Google’s announcement does not establish how the system performs across every Mac application, or how reliably it recovers when a workflow encounters an unexpected screen.
The company’s privacy and safety descriptions are therefore claims from the announcement rather than independent test results. There is no hands-on testing behind this report, and Google has not said that the Mac feature is available on every account or in every region.
Availability and Mac requirements
Computer use is rolling out in English in the United States through the Gemini app for Mac. Google says it requires an Apple Silicon Mac running macOS Sequoia 15.0 or later, a personal Google account for users aged 18 or over, and a Google AI Plus, Pro or Ultra subscription.
Google directs users to Gemini’s Mac download page to try the feature. The company has not announced wider regional availability or a release date for other platforms in the announcement, so readers outside the United States should not assume the rollout is live for them yet.
Our opinion
This is a more useful direction for desktop AI than another promise of a smarter chat window. The interesting part is not that Gemini can click buttons; it is that Google is packaging those actions around a visible, interruptible workflow. A live preview and app-by-app permissions give users a chance to catch a mistake before it becomes a completed form, altered file or published listing.
The limitation is equally clear. Google’s examples are polished demonstrations, and the announcement does not answer the practical questions that decide whether computer use is genuinely dependable: how often it gets stuck, what happens when an app changes its layout, how much supervision each task needs, and whether the permission model remains understandable once several apps are involved. Until those questions are tested outside the launch examples, Gemini computer use looks promising for supervised chores rather than a reason to let an AI loose on an unattended Mac.
For readers who already pay for an eligible Gemini plan and regularly repeat the same cross-app tasks, the feature could save time. For everyone else, the United States-only rollout and Apple Silicon requirement make this an announcement to watch rather than an immediate upgrade.