Key takeaways
- Google said on Nov. 19 that the Gemini Mac app can now carry out multistep tasks across local files, desktop apps and websites by clicking, typing and filling in forms.
- Gemini runs these tasks in an isolated background window with a live preview in the chat, and users choose which apps it can reach, can pause it and must approve key actions.
- The feature is rolling out in English in the U.S. to Google AI Plus, Pro and Ultra subscribers age 18 and older on Apple silicon Macs running macOS Sequoia 15.0 or later.
In this story 4 sections
Google has added computer use to the Gemini Mac app, letting its AI assistant click, type and fill in forms inside desktop programs and websites on a user's behalf. The update was announced Thursday, Nov. 19, in a Google blog post by Michael Friedman, a group product manager for the Gemini app at Alphabet (NASDAQ: GOOGL) unit Google.
Until now, the Gemini app on the Mac could answer questions and work with what a user shared in the chat. With this change, a user describes a multistep chore in plain language and Gemini carries it out across local files, installed apps and the web. Google pitched it as a way to skip long runs of repetitive clicking.
How Gemini works on the Mac without taking over the screen
The agent does its work out of sight of the main desktop. "Gemini works inside an isolated background window with a live preview right in your chat," Friedman wrote. According to the post, that means the user's own cursor and screen stay free, so they can keep working in other apps while the task runs.
Google listed several controls meant to keep people in charge of what the assistant does:
- The user picks which apps Gemini is allowed to reach.
- Each step can be watched as it happens in the chat preview.
- A task can be paused at any point.
- Gemini asks for approval before it takes key actions.
The post does not describe the permission screens in more detail, and it does not say what kinds of actions count as key ones.
What Google says the feature can do
The company gave four sample jobs to show the range it has in mind:
- Applying the same color grade to a batch of photos in design software, or building a detailed 3D scene.
- Reading weights, sizes and addresses out of PDF files saved on the Mac and entering them into shipping quote forms with many fields.
- Gathering figures and screenshots from live dashboards for several regions and turning them into a Google Slides presentation.
- Looking at photos of clothing in a folder, checking prices online, recording item details and preparing listings on resale sites.
"Just describe what you need, and Gemini handles the clicking, typing, menus, and forms for you," Friedman wrote.
Who can use Gemini computer use on a Mac
Access is limited at launch. Based on the blog post and its footnotes, a user needs all of the following:
- A paid Google AI Plus, Pro or Ultra subscription.
- A personal Google account held by someone 18 or older.
- A Mac from Apple (NASDAQ: AAPL) with Apple silicon running macOS Sequoia 15.0 or later.
- English as the language, with the rollout covering the U.S. only.
Google did not give a timeline for other countries, other languages or work accounts, and it did not list a separate price for the feature beyond the subscription tiers. The app is offered as a download from Google's gemini.google/mac page, the post said.
Where this fits in Google's Gemini rollout
The Mac update is the latest in a run of Gemini releases this month. Google recently opened its Antigravity Teamwork tools to developers using Gemini 4 and launched Gemini 4 Neon with published benchmarks and pricing. Computer use on the desktop moves the assistant from answering in a chat window toward operating software directly, and Google has paired that shift with the approval steps described above.
For more coverage of AI assistants and models, see our AI news section.
Share this story
Spotted an error? Tell us


