What problem does it solve?
It removes the friction of performing GUI tasks manually by letting an AI interact with your Mac’s desktop—clicking, typing, and navigating—without interrupting your active work.
Core Features & Use Cases
- Background desktop control: Takes screenshots, clicks, types, and scrolls without stealing cursor position, keyboard focus, or Spaces.
- Model-agnostic operation: Works with any tool-capable model by relying on a capture/click/verify loop instead of model-specific UI schemas.
- Element-index reliability: Uses AX-tree indices and numbered overlays so the AI can reference UI elements consistently.
- Use case: Have the AI complete a multi-step workflow in Safari (e.g., searching, filling a form, and confirming results) while you keep typing in your editor.
Quick Start
Ask: “Use the computer use tool to capture the current Safari screen, click the element that starts the sign-in flow, and then re-capture to confirm what page I’m on.”