Your Mac just got a new kind of helper — one that does not just answer questions, but actually does the clicking for you. On November 19, 2026, Google announced that its Gemini app for Mac can now use your computer on your behalf, navigating your apps, files and websites in the background to finish multi-step tasks while you keep working on something else.
It is called Computer Use, and it is Google’s biggest step yet toward turning Gemini from a chatbot into a working agent. Instead of telling you how to do something, Gemini can now do it: open an app, work through the menus, type into the fields and hand you the finished result. The timing is no accident — the race to put an AI agent on your desktop is now in full swing, and Google does not intend to watch from the sidelines.
What Is Gemini Computer Use?
Computer Use is a new capability inside the Gemini app for Mac. You describe a task in plain language, add “@Computer use” to your prompt, and Gemini takes over the clicking, typing, menus and forms for you. It works across your local files, your desktop apps and the web, so a single request can span several different tools — pulling numbers from a live dashboard, for example, and turning them into a presentation.
Google’s announcement came on its official blog, with Gemini group product manager Michael Friedman explaining that the goal is the repetitive work that normally takes dozens of clicks. The idea is simple: the boring middle of a task disappears, and you only see the result. TechCrunch’s report on the launch provides further detail on how the feature works.
It is worth noting that Google is not first to this idea. Anthropic’s Claude has offered computer use for over a year, and Meta’s recently launched Muse agent for Mac can also work with local apps and data. What is new here is that Gemini — already sitting inside Google’s apps, search and phones — can now reach into the desktop itself.
What It Can Actually Do
Google’s examples are deliberately practical, aimed at the chores that eat up an afternoon. Here is what the company says Computer Use can handle today:
- Speed up creative workflows. Batch color-grade photos in your favourite design software, or set up complex 3D scenes, without clicking through the same panels fifty times.
- Skip manual form entry. Pull weights, dimensions and addresses from local PDFs and use them to fill out multi-box shipping quotes.
- Turn live data into ready-to-present decks. Grab metrics and screenshots from live dashboards across several regions and build a presentation in Google Slides.
- Batch-list items for sale. Analyse clothing photos from a local folder, research pricing, log the item details and stage listings on resale platforms.
Under the hood, Gemini can handle tasks made up of dozens of actions, and the output of one workflow can feed the next — generate a visual asset in a design tool, for instance, and then turn it into a small app with Visual Studio Code. It can also move data in both directions between local apps and web services, loading information from web pages into programs on your machine and back again.

The feature also plugs into Gemini’s existing connectors. In the app’s Customize menu, new “Desktop” apps such as Blender and Autodesk Fusion now sit alongside the existing Git and Obsidian integrations, so Gemini can reach into the tools creative and technical professionals actually use. 9to5Google’s breakdown of the launch walks through the feature’s capabilities in more detail.
How It Works Without Taking Over Your Screen
The cleverest part of Google’s design is where the work happens. Gemini does not grab your mouse or lock your screen. Instead, it works inside an isolated background window, with a live preview right inside your chat. Your main screen and cursor stay completely free, so you can keep typing, designing or taking calls while the agent works beside you.
That is a meaningful difference from the first generation of computer-use tools, which often took over the whole screen and left you watching the AI click around. Google’s approach treats the agent more like a colleague working at the next desk: visible, but out of your way. You can watch every step live in the chat preview, send feedback mid-task, and pause or stop the agent at any time.
You Stay in Control
Handing an AI the keys to your computer is a big ask, and Google knows it. The company says privacy and safety are built in, and the controls are all in your hands:
- You choose which apps Gemini can access. Nothing is open by default — you decide what the agent is allowed to touch.
- You watch every step live. The preview in your chat shows exactly what Gemini is doing as it does it.
- You can pause at any time. One tap stops the agent mid-task.
- You approve key actions before they happen. Gemini asks first before doing anything significant.
That last point matters more than it sounds. Google’s own support documentation warns that AI agents can run into misleading instructions hidden inside apps and websites — a prompt-injection risk that every computer-use tool shares. Careful permission management is not a nice-to-have here; it is the whole safety model. And it is a concern users are already acting on: Signal’s new screenshot blocking exists precisely because people do not want AI agents reading their private chats.

There is also a telling bit of timing. Apple’s upcoming macOS 27.3 update is expected to warn users more clearly about the dangers of giving apps unrestricted access to their data, and to let people limit agents to specific apps and folders. The operating system itself is about to get stricter about exactly the kind of access Gemini is now asking for — which makes Google’s permission-first pitch look well judged.
Who Gets It — and What You Need
Computer Use is rolling out now, but not everyone gets it on day one. At launch it is available in English in the United States, for personal Google accounts aged 18 and over, on a Google AI Plus, Pro or Ultra subscription. Google AI Plus starts at $4.99 a month, so the cheapest tier does qualify — but free Gemini users are left out for now.
On the hardware side, you need an Apple Silicon Mac running macOS Sequoia 15.0 or later. Intel Macs are not supported. The Gemini app for Mac is available to download from Google’s own site, and the feature appears once the rollout reaches your account. MacRumors also covered the launch, including the Apple Silicon and macOS requirements.
That narrow launch is worth keeping in mind if you go looking for it today and do not see it. Rollouts like this usually take days or weeks to reach every eligible user, and Google has not given a date for wider availability — including other countries or the free tier.
The Bigger Race: AI Agents Are Moving Onto Your Desktop
Gemini’s arrival on the Mac is one move in a much bigger shift. The AI industry spent 2025 teaching chatbots to answer questions; in 2026 it is teaching them to do things. This same week, Google also widened the rollout of Gemini Tasks, which lets the assistant sign in to websites, make purchases and book appointments on your behalf, and it has begun admitting early testers to a revamped family-organising agent first announced in September.
The competition is moving just as fast. Apple’s own AI agent for the Mac is reportedly in development, designed to handle complex requests across applications, browsers and external services. OpenAI’s delayed AI device shows how hard the hardware side of this race is, while the software side — agents that live on the machines people already own — is moving much quicker.
Underneath it all is a more powerful model, too. Gemini 4 Neon, launched just a day before Computer Use, brings frontier-level reasoning to complex, high-frequency workflows at flash speeds — exactly the kind of model an agent needs to plan and execute dozens of steps without going off track.
Meanwhile, Google is reshaping its own product line around this agent-first future, retiring Google Assistant on home devices as Gemini takes over. The assistant you talk to is becoming the agent that acts for you, on every screen you own.
The Bottom Line
Computer Use turns Gemini on the Mac from an advisor into a doer. The background-window design is genuinely thoughtful — you get the help without losing your screen — and the permission controls are the right starting point for a tool that can click and type on your behalf. The catch is the gate: a paid subscription, an Apple Silicon Mac and a US rollout mean most people will be watching from the sidelines for now.
But the direction is unmistakable. Within a year, having an AI that can operate your computer will probably feel as normal as having one that can answer your questions. Google has just made sure Gemini is part of that future — and the next question is not whether these agents work, but how much of your computer you are willing to let them touch.

