computer-use - Control desktop applications in the background
Drive the desktop across macOS, Windows, and Linux without moving the user's cursor or stealing keyboard focus.
Tags
Updated: 2026-10-05Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Capture desktop state
- Click interface elements
- Type text input
- Send keyboard shortcuts
- Drag interface elements
- Scroll application views
- List running applications
- Focus desktop applications
- Verify action effects
- Escalate input delivery
Inputs
- Target desktop application
- Action parameters
- Element indexes
- Screen coordinates
- Browser window identifiers
- Browser tab identifiers
Outputs
- Screenshots
- Accessibility trees
- Action verdicts
- Application state changes
- Browser state snapshots
Requirements
- macOS, Windows, or Linux
- computer_use tool
- Tool-capable model
- cua-driver runtime