What problem does it solve?
This Skill removes the gap between “AI understanding” and “real UI execution” by letting agents directly operate native desktop apps, Chrome/Electron, and Android devices through a local MCP server.
Core Features & Use Cases
- Multi-modal UI control: Screenshots + OCR/text search and click/type/drag/scroll actions to automate apps even without accessibility hooks.
- Accessibility-first native automation (macOS): AX dispatch to click and set values using an accessibility tree for precise, cursor-free control.
- Web automation for Chrome/Electron (CDP): DOM-level control for finding elements, filling forms, waiting on conditions, and inspecting UI.
- Android automation (ADB): Screenshot, text lookup, tapping, typing, and app launch/stop via connected devices.
- Use Case: Automate a macOS settings flow by taking an AX snapshot, clicking the right controls, and verifying the resulting state via a new snapshot or screenshot.
Quick Start
Run the MCP server, then connect your MCP-compatible client and ask it to automate your target desktop app, browser, or Android workflow.