The computer use skill
for Google Antigravity

Type /computer-use and your agent sees the screen, reads each app's accessibility tree and drives real apps on Windows and macOS with your mouse and keyboard. Esc always stops it.

/computer-use let Gemini 4 Hydrogen paint this in Microsoft Paint. One prompt in Antigravity. The agent drove Paint through a stroke plan scripted from the photo, shown 200× faster. The bar and glow are drawn in, since they're hidden from screen recordings by default.

2,553real mouse strokes
40exact colours
~65 minunattended
1prompt

Look closer

Hover over the painting to see the strokes. Every one was a real drag of the mouse in Paint.

The portrait Gemini 4 Hydrogen painted in Microsoft Paint
Painted through /computer-use
The original photo
Original photo

On a phone, tap an image to open it full size.

The painting at 9, 24 and 43 minutes, and finished at about 65 minutes

How it works

One skill and a small local service. It uses the model you already chose in Antigravity: no API key, no second LLM, no MCP server.

Observe

Reads the window's accessibility tree, the names, roles and values of its controls, and takes a screenshot only when a task needs eyes.

Act

Clicks, types, drags and fills in dialogs with the real mouse and keyboard, on the elements it just observed.

Verify

Checks a fresh view after each step. An input whose outcome is uncertain is never sent twice.

What it can do

Use native controls

Click, select, toggle and set values on buttons, fields and lists by name.

Work visually

Click, double-click, right-click, scroll and draw strokes through hundreds of points on a canvas.

Type anything

Unicode text, shortcuts, and paste that puts your clipboard back afterwards.

Work across apps

Launch, find and focus windows, including each app's own file dialogs.

Run long jobs

plan sends many steps at once, so a 30-turn task drops to about 9.

Update safely

Signed releases, automatic updates you can switch to manual, and rollback.

You stay in control

See what it's doing

A small bar always says what the agent is doing right now, and it changes the moment the state does.

Antigravity is working

While it works

A glow on the window it's using

Soft light around the window the agent is working in, so you always know where it is. It's left out of the agent's own screenshots.

A window lit by the soft blue and violet edge glow

It asks before touching an app

Nothing runs without your OK. Terminals and code editors are click-only, so it can't type commands into them.

An access card asking to allow Antigravity to use Notepad

One key stops it

Esc or Ctrl Alt F12 stops it at once and lets go of every key and button. Sign-ins and admin prompts are always left to you.

The Settings menu with the stop keys

Proven on real apps

Not a sandbox and not a browser extension: the agent works in the same desktop apps you do.

Microsoft Paint

The portrait above: exact colours set through Paint's own colour editor, then painted as long strokes with the real mouse.

2,832steps in one plan
~65 minunattended

Upscaly

Open the app, pick a photo in its file dialog and run a 4× upscale. Batching steps cut the turns the model needs.

30 → 9model turns
4.5 → 2 minend to end

Private and careful by design

Your data stays yours

  • No telemetry. Screenshots go only to the model you already use in Antigravity.
  • The glow and bar are left out of the agent's own screenshots.
  • Sign-ins, admin prompts and the secure desktop are always handed back to you.

Safe to keep running

  • Every release is signed, and installed copies update only from signed metadata.
  • Updates can be switched to manual and rolled back.
  • An input whose outcome is uncertain is never replayed.

Platforms

Windows

Windows 10 and 11, x64. One-line install in PowerShell; uses the .NET Framework already on Windows.

macOS

macOS 14 or later on Apple silicon or Intel. You grant Accessibility and Screen Recording once.

Questions

Does it need its own API key or model?

No. It uses the model you already chose in Antigravity. There's no second LLM, no API key and no MCP server to set up.

Does it work with agents other than Antigravity?

It is built and tested for Google Antigravity. The service is driven by a plain command-line tool, so other agents that can run shell commands can call it, but only Antigravity is supported today.

What stops it from doing something I don't want?

It only acts in apps you approve, terminals and code editors are click-only, and Esc stops it at once and releases every key and button. Your own mouse and keyboard never pause it, so stopping is always one key away.

Is it fast?

Most of the time in a long task is the model thinking between steps. plan removes much of that by sending many steps in one command.

Is it open source?

Yes, MIT licensed. It is an independent project, not affiliated with or endorsed by Google.

Install

Windows, in PowerShell

irm https://raw.githubusercontent.com/shayann07/antigravity-computer-use/main/install.ps1 | iex

macOS (Apple silicon or Intel)

Download the macos-arm64 or macos-x64 archive from the latest release, then run install.command and Setup.Mac.Permissions.command. Mac guide

Then try

/computer-use
Open Paint, draw a red circle in the middle, save it to my Desktop as circle.png.