"author:freehul" matching MCP servers:

Enables AI agents to inspect and drive Windows desktop applications text-first through the UI Automation tree, reading interactive elements and performing element-level actions (click, invoke, set value, type, drag, menu and window/app lifecycle) without relying on screenshots or coordinates. Actions are dry-run by default and pass an internal confirmation gate with snapshot binding and an audit log, while deterministic waits/verifications and an explicit screenshot or OCR fallback cover self-drawn surfaces.