macOS Native GUI Automation (CUA Driver)

•
20
•

Drives native macOS applications without stealing focus by utilizing virtual agent pointers, accessibility trees, and MCP workflows.

The CUA Driver skill enables Claude Code to interact with native macOS desktop applications seamlessly in the background without stealing user focus or hijacking the physical cursor. By leveraging OneWorks Cua Driver MCP tools, accessibility APIs (AX), and coordinate-based fallbacks, Claude can inspect window states, click, type, scroll, and orchestrate complex multi-app workflows while maintaining a clear visual evidence trail with distinct agent pointers.

Key Features

01Accessibility tree inspection (AX) with coordinate fallback for custom canvas apps
02Multi-step and multi-app orchestration via workflow execution tools
03Non-invasive window inspection and screenshot verification
04Virtual agent pointer rendering with customizable session colors and positions
05Background app interaction without hijacking the user's cursor or focus
0620 GitHub stars

Use Cases

01Automating repetitive macOS desktop tasks without interrupting active work
02Running end-to-end GUI verification tests on native desktop and Electron apps
03Extracting data and verifying visual states across desktop applications