Desktop automation for AI agents — screenshot, click, type, scroll, accessibility trees, form filling, menu navigation, scripts (AppleScript/JXA/PowerShell), app/window management. 49 tools.
| Workflow | Calls | What It Achieves |
|---|---|---|
| See → Click → Verify | 3 | Screenshot, interact, confirm |
| Fill Form | 2 | Find fields → fill all |
| Menu Action | 2 | Find shortcut → press key |
| Script Automation | 1 | Complex AppleScript/PowerShell |
| App Management | 2 | List → activate |
- Keyboard shortcut — fastest (1 call)
- Accessibility click — most reliable (click_element)
- Script — most powerful (run_script)
- Coordinate click — fallback (screenshot + left_click)
git clone https://github.com/zavora-ai/skill-computer-use.git \
~/.skills/skills/computer-useRequired: computer-use-mcp (49 tools)
Platforms: macOS (AppleScript/JXA), Windows (PowerShell)
Permissions: Accessibility access in System Settings
Cross-MCP:
mcp-browser— web automationworksheet-mcp— open generated reportsmcp-slack— desktop notification context
| Metric | Target |
|---|---|
| Click accuracy | Correct element on first attempt |
| Safety | Never click without screenshot |
| Approach | Accessibility over coordinates |
![]() James Karanja Maina |
|---|
Apache-2.0 — Part of ADK-Rust Enterprise. Built with ❤️ by Zavora AI
