Google teaches Gemini to run your Mac in the background without taking over the screen or cursor
Google has added a computer use mode to the Gemini app for macOS. The AI opens apps on the Mac, browses websites and completes multistep tasks on the user's behalf. While the agent works, it shows a live preview in the chat, and the screen and cursor stay free. Any action can be paused, stopped or approved in advance. Bottom line: you can hand forms and folder cleanup to Gemini, but for now only in the US, in English and with a paid subscription.
AI-processed from TechCrunch; edited by Hamidun News
Imagine you need to sort a hundred files into folders, fill out the same form for a dozen clients, or build a presentation from figures in an online dashboard. It is tedious work that nobody wants to do by hand, but it is not worth the effort of writing a script either. Google now suggests handing exactly these tasks to Gemini: the desktop app for macOS has a new computer use mode, meaning the AI operates the computer in place of the human.
What exactly Gemini can do on Mac
Google announced the feature on Thursday. The agent can navigate native macOS apps and open websites in the background. You describe the task in plain text, and Gemini then takes on a multistep job that touches local files, apps and the web at the same time.
The key convenience: while the agent works, the chat shows a live preview of what it is doing. The screen and cursor stay with the user. You don't have to sit and watch the mouse move across the desktop on its own; you can calmly get on with something else.
Controls are in place too. As with other AI agents that operate a computer, you can pause Gemini's actions, stop them entirely or approve them before they are carried out. For a feature with access to files and websites, this is a basic requirement: mistakes in these scenarios cost more than they do in an ordinary chat.
Google names these use cases:
- filling out forms;
- repeating one action across a batch of files;
- sorting files into folders by given criteria;
- creating slides from files or data in online dashboards.
Google catches up with competitors
TechCrunch notes that the company is arriving late: OpenAI, Anthropic and Meta have been offering agentic features in their desktop apps for some time. For Google, this is another step in its agent lineup. This week it expanded access to the "Tasks" feature in Gemini, where the AI helps log in to websites, make online purchases and book appointments. The company has also just begun admitting early testers to an updated agent called "CC", announced in September. It is aimed at families and is meant to help tidy up email, calendars, chats and to-dos.
The pieces now fit together: Tasks works on the web, CC handles family organization, and computer use on Mac goes beyond the browser to reach local files and apps. For product managers, this is a signal that the competition is no longer about the best chatbot, but about who becomes the operating layer between the user and their computer.
Limitations and what to do about them
The launch is narrow for now. The feature is available in English and only in the US. You need the Gemini app for Mac and a Google AI Plus, Pro or Ultra subscription. The company has said nothing about timing for expansion to other markets and platforms, such as Windows. The announcement gives no details on how the agent handles errors, how reliable it is on complex apps, or what system permissions it needs, so it is too early to judge quality.
The practical takeaway for developers and founders is this. If you are building your own agents, expect users to compare them with the built-in offerings from Google, OpenAI and Anthropic: background operation that doesn't hijack the cursor and explicit confirmation of actions are becoming the expected minimum. If you simply use a Mac and a Google subscription, it makes sense to start with safe routine tasks such as sorting files and drafting forms, and to keep action confirmation turned on until you understand how the agent behaves with your data.
- Nov 20, 2026Google Built a Sealed-Off Copy of the Internet to Test Chrome's AI Agents
- Nov 20, 2026Anthropic shares jump 17% on first day after record $81.3 billion IPO
- Nov 20, 2026Google opens Genie 4 world model in its API: one minute of walking costs about $6
- Nov 20, 2026Gemini API Cuts Token Use for a 200-Tool Agent Nearly 7x, by Google's Estimate
Key Questions Answered
What exactly Gemini can do on Mac?
Google announced the feature on Thursday. The agent can navigate native macOS apps and open websites in the background. You describe the task in plain text, and Gemini then takes on a multistep job that touches local files, apps and the web at the same time. The key convenience: while the agent works, the chat shows a live preview of what it is doing. The screen and cursor stay with the user. You don't have to sit and watch the mouse move across the desktop on its own; you can calmly get on with something else.
What are the key details on Google catches up with competitors?
TechCrunch notes that the company is arriving late: OpenAI, Anthropic and Meta have been offering agentic features in their desktop apps for some time. For Google, this is another step in its agent lineup. This week it expanded access to the "Tasks" feature in Gemini, where the AI helps log in to websites, make online purchases and book appointments. The company has also just begun admitting early testers to an updated agent called "CC", announced in September.
What are the key details on Limitations and what to do about them?
The launch is narrow for now. The feature is available in English and only in the US. You need the Gemini app for Mac and a Google AI Plus, Pro or Ultra subscription. The company has said nothing about timing for expansion to other markets and platforms, such as Windows. The announcement gives no details on how the agent handles errors, how reliable it is on complex apps, or what system permissions it needs, so it is too early to judge quality. The practical takeaway for developers and founders is this.
Build & Ship Production AI Agents Without Boilerplate Code
Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.