Google adds computer use to the Gemini app for Mac, letting it operate local files, desktop apps and websites in a background window
Edited by Elliot Marsh, Editor · written with AI from the linked sources (how)
Starting 19 Nov 2026, the Gemini app for Mac has a computer use feature. It carries out multi-step tasks by clicking, typing, navigating menus and filling forms across local files, desktop apps and the web. It runs in an isolated background window with a live preview in the chat, so the user's cursor and screen stay free. Users choose which apps Gemini can access, can pause at any time, and approve key actions before they happen. It is rolling out in English in the U.S. only. It requires an Apple Silicon Mac on macOS Sequoia 15.0+ and a personal Google account (18+) with Google AI Plus, Pro or Ultra.
Entities: Google, Gemini app for Mac, Gemini, Apple macOS, Google Slides, Google AI Plus, Pro and Ultra
1 primary
What happened
From 19 Nov 2026, Google's Gemini app for Mac has a "computer use" feature. It can click, type, navigate menus and fill in forms across local files, desktop apps and websites, working in an isolated background window with a live preview in the chat. Users pick which apps Gemini may access, can pause it, and must approve key actions. It is limited to English in the U.S., Apple Silicon Macs on macOS Sequoia 15.0 or later, and personal Google accounts (18+) on a paid Google AI Plus, Pro or Ultra plan.
Why it matters
This moves desktop-controlling agents from developer tools and demos into a mainstream consumer app, which is a real distribution step. The people affected first are paying Google AI subscribers in the U.S. who do repetitive admin work, such as form filling, batch edits and listing items for sale. The background-window design and approval prompts are sensible, but they also show how much trust the feature needs. For competitors it adds pressure to match on ease of use and safety controls, not on raw capability. Real impact is limited for now by the narrow rollout and paid-tier gating.
What is noise
The "dozens of clicks" and time-saving claims are marketing examples with no published accuracy, speed or failure-rate data. Similar computer-use agents already exist from other labs, so this is closer to catch-up and packaging than a breakthrough. Nothing yet shows how reliably it handles messy real-world apps or how well the approval prompts stop mistakes.
Watch next
- 01Independent hands-on tests over the next few weeks: task completion rates, error rates, and whether approval prompts catch bad actions on real multi-step jobs.
- 02Expansion beyond U.S. English, Windows, or lower subscription tiers, and any Google statement on timing, as a sign of how confident it is in the feature.
- 03Security reports of prompt injection or unintended actions via web pages or files, and how Google and rival agents respond.
Coverage
4 stories- Let Gemini tackle repetitive chores on your MacGoogle Gemini Blog · primary · 19 Nov 2026Tier 1
- Gemini on Mac can now use your computerThe Verge AI · 19 Nov 2026Tier 2
- Google gives Gemini the ability to use your Mac for youTechCrunch AI · 19 Nov 2026Tier 2
- Google's Gemini can now run Mac apps and finish desktop chores in the background while you keep workingThe Decoder · 20 Nov 2026Tier 2
Newsletter
Get the no-hype AI news straight to your inbox
Weekly. No spam, unsubscribe any time. Privacy
More capability signals
Full feed →- Deepseek releases V4.1-Flash, an open-source model that sharply cuts KV cache memory and input-processing compute for AI agents10 Sept 202682
- Mistral releases Mistral Large 4, a ~1T-parameter open-weight multimodal MoE model, with day-0 Hugging Face support29 Oct 202680
- Meta releases Muse Spark Open, an open-weight ~816B-parameter multimodal mixture-of-experts model, with Hugging Face support15 Oct 202680
- OpenAI discloses sandbox-escape and credential-leak incidents, confirms pause on tool-use for its most capable models26 Sept 202680