Google Brings Computer Use To Gemini On The Mac

Google on Thursday began rolling out computer use in the Gemini app for macOS. The feature lets the assistant click, type, and navigate menus inside native Mac apps and websites to complete multi-step tasks, rather than just describing how a user would do them.
How it works
Users invoke the agent by adding "@Computer use" to a prompt and describing the workflow. Gemini then operates in an isolated background window with a live preview shown in the chat; the user's own screen and cursor stay free for other work. Google's example uses are the tedious ones: filling out forms, repeating one operation across a batch of files, sorting folders by criteria, or building a slideshow from dashboard data.
Control is layered. Users choose which apps Gemini may access, can watch and pause its progress, and key actions require approval before they happen; Google says it pauses for logins and payments. With permission, Gemini can view and edit most user folders, including Downloads, Desktop, Documents, and Photos, so which folders are connected determines what local data enters a task. The rollout also adds desktop app connectors, including Blender and Autodesk Fusion, alongside existing Git and Obsidian integrations.
Late, but bundled
Google is the last major model lab to ship desktop computer use. OpenAI, Anthropic, and Meta have offered agentic features in their desktop applications for some time, and Google's own Tasks feature (formerly Gemini Spark) already worked across local files and connected Google apps on the Mac.
Requirements are narrow for now: Apple Silicon running macOS Sequoia 15 or later, an English-language account in the US, and a Google AI Plus, Pro, or Ultra subscription. AppleInsider noted that Google's announcement lists all three tiers while its support documentation initially specified only Pro and Ultra. Google did not say when the feature would expand to other markets or platforms.
The subscription bundling is the commercial point. Computer use is not sold as a separate product; it is an argument for moving a workflow into the Gemini app, the same distribution play Google ran with Gemini in Chrome. Whether users hand a production Mac an agent that can edit their Documents folder comes down to how much they trust the approval flow, and Google has clearly designed the isolated window and per-action consent to be the answer to that objection.
Sources:
