computer-use-rs is a cross-platform Rust crate that gives your code the same basic abilities as a person at a computer. It can look at any desktop app, read what's on screen, and click, type, scroll and drag in it.

We built it to power Computer Use in the AgentOne desktop app, and now we're open-sourcing it for everyone.

What you get

• See an app: get its accessibility tree as an indexed list of elements, a PNG screenshot of its window, and a plain-text outline, all from one call.

• Act on it: click elements by index or screenshot pixel, type text, press key chords like ctrl+shift+Tab, scroll, drag, set values, select text, and trigger named actions like "Show Menu".

• Work in the background: by default, input goes to the target app without moving your mouse or stealing focus, wherever the OS allows it.

• Runs on three platforms: macOS (Accessibility API + ScreenCaptureKit), Windows (UI Automation), and Linux (AT-SPI + X11).

• Just a library: a typed Rust API with no server, runtime or protocol attached. Computer is Send, so you can use it from any thread.

• Easy to test: the platform layer is a Backend trait, so you can drive a fake desktop in tests or add support for a new platform.

Use it for: AI agents that operate a computer, UI automation and scripting, end-to-end tests of desktop apps, and accessibility tools.

`cargo add computer-use-rs`. Dual-licensed under MIT or Apache-2.0.