What problem does it solve? Automating native Windows applications, external browser windows, and OS dialogs requires reliable UI observation and input delivery that shell commands and web tools cannot provide. ## Core Features & Use Cases - Window observation and capture: Inspect windows via accessibility trees, OCR marks, screenshots, and zoomed frame regions before acting. - Structured input actions: Click, type, scroll, drag, and send keys using semantic refs, with background or foreground delivery modes. - Verification and recovery: Use bounded verify predicates, diagnose backend readiness, and follow strict pause/resume and cleanup flows after interruptions. - Use Case: Ask the agent to fill a form in an external Chrome window or operate a native Windows settings dialog while preserving the user's session and focus. ## Quick Start Use the computer tool to capture the external Chrome window and click the submit button in the open form.