Gemini for Mac can now operate your apps, working in a window of its own
Google's computer use feature clicks, types and fills forms in the background, but only for paying US subscribers on Apple silicon Macs.

Google's Gemini app for Mac can now operate other software for you. A new computer use feature lets the assistant click buttons, type, work through menus and fill out forms in desktop apps and on websites. It does that work in a separate background window, so your own screen and cursor stay free.
Google announced the feature Thursday and began rolling it out in English in the U.S. the same day. It is limited to paying customers on Google AI Plus, Pro and Ultra plans, which start at $4.99 a month. Users must be 18 or older and signed in with a personal Google account. Hardware is restricted too: the feature needs an Apple silicon Mac running macOS Sequoia 15.0 or later.
The background window is the most practical part of the design. Many computer-use agents work by taking over the screen, which turns a time-saving tool into something you have to sit and watch. Google's version shows its progress as a live preview inside the Gemini chat. You can check in on it while you keep working.
What the agent can actually do
Until now, the Mac app's agent features were mostly about files. The capability Google first called Gemini Spark, now known as Tasks, could read and reorganize local files. It could also pull from connected services such as Gmail and Google Calendar to sort folders or turn documents into spreadsheets. Operating the apps on the machine is a bigger step. In Google's words, the goal is to help users "get multi-step tasks done across your local files, desktop apps, and the web."
Google's examples are mostly the dull, repetitive work that eats up an afternoon:
- batch-editing a folder of photos
- preparing shipping quotes from PDFs
- creating listings for resale platforms
- sorting files into folders based on set criteria
- building presentations from metrics, screenshots or data spread across several websites and online dashboards
The company's showcase demo aims higher. A parent asks Gemini to turn a child's drawing of a butterfly into an animated 3D model in Blender, then bring it to life in an augmented reality web app. That kind of workflow is where computer use could matter most. Blender is powerful but intimidating to newcomers, and an assistant that can drive it from a plain-language request could open it to people who would never learn its interface. Whether Gemini handles that reliably outside a polished demo is the real test, and the launch materials don't settle it.
To start a job, users type @Computer use in a prompt and describe the result they want. Under Customize, then Connectors, the app now lists "Desktop" apps such as Blender and Autodesk Fusion. They appear alongside existing integrations with Git and Obsidian.
Guardrails, and Apple's coming restrictions
Google has built in several limits. Users pick which apps Gemini may touch, and a single click stops the agent. Gemini pauses when it reaches a login or payment screen and asks for approval before what Google calls "key actions." According to Google, the agent runs in an "isolated background window with a live preview right in your chat."
Those controls look sensible. Their usefulness, though, depends on how Google defines a sensitive action, and the company's announcement doesn't spell out where that line falls. An agent that asks too often gets tedious. One that asks too rarely can submit the wrong form or overwrite the wrong file.
The launch also lands just before a change on Apple's side. In macOS 27.3, Apple plans to tighten rules for apps that want full disk access, citing the risks posed by AI agents with computer use. A preview of the new settings appeared in this week's macOS 27.3 beta. Apple will warn users about giving apps unrestricted access to their data and let them limit access to particular apps and folders. Apps can also be set to request permission every time. A new indicator will show when one app is controlling another.
MacRumors reported that apps granted Mac permissions before macOS 27.3 arrives won't have to ask again after the update. Users who already granted them also won't see Apple's new explanation screens. In practice, then, people who set up Gemini's computer use now will skip Apple's extra warnings. Apple's own tools for narrowing that access will still be available once the update ships.
For anyone handing an AI agent the run of their Mac, those Apple controls are worth revisiting when they arrive. Two independent layers of permission, one from Google and one from the operating system, are better than relying on the agent maker's settings alone.
A busy week for Google's agents
Computer use caps a run of agent announcements from Google. Two days earlier, the company made Gemini Tasks free. It also gave the assistant the ability to call businesses, make purchases, and take requests by text and email. TechCrunch noted that Tasks can also sign into websites and book appointments. Google has also started letting early testers try a revamped agent called CC, announced in September. CC is designed to help families keep track of email, calendars, chats and to-dos.
Android Authority reported in September that Google was testing computer use, so the feature itself isn't a surprise. The pricing choice stands out more. Google opened Tasks to everyone this week but put computer use behind a subscription, making desktop control the premium rung of its agent lineup. A $4.99 entry plan keeps the price of trying it low. Free users, though, get a taste of agents that stops short of the most capable one.
The release also leaves Google playing catch-up. TechCrunch pointed out that OpenAI, Anthropic and Meta have offered agent features in their desktop apps for some time. Google's main advantage here may be reach. Gemini is already wired into Gmail, Calendar and the rest of Google's services, so an agent that can also operate local apps can link up tasks that competitors may need extra setup to connect.
The platform choice is notable too. Launching on the Mac first, rather than on Windows or on Google's own ChromeOS, puts Google's most ambitious desktop agent on a rival's operating system. It also makes Apple's permission rules a direct factor in how the feature works. Google didn't say when computer use would come to other countries, other languages or other platforms.
The Gemini app for Mac is available to download from Google's website.
Sources
- Android Authority — Android News: Google brings computer use to Mac, letting Gemini handle chores without hijacking your screen
- MacRumors — Apple News: Google's Gemini Can Now Control Apps on Your Mac
- 9to5Mac — Apple News: Gemini app for Mac adds computer use to control desktop apps
- The Verge — Tech: Gemini on Mac can now use your computer
- TechCrunch — AI: Google gives Gemini the ability to use your Mac for you




