Docs

Requirements: Windows, Python 3.11+ (3.12 recommended). Microsoft Office and Tesseract OCR are optional.

Install: pip install -r requirements-lite.txt

Run: python run_agent.py — then type a natural-language command or press “/” to open the command palette.

The app moves your real mouse and keyboard — keep a hand near Esc the first few times.

Type “/” in the input to open the palette. Commands are grouped into Cursor, Mouse and Keyboard.

Arrow keys highlight, Enter runs, Tab inserts the expansion, Esc closes.

Examples: /copy → press ctrl+c · /move-center → move cursor to center · /save → press ctrl+s.

Simple → Enhanced (pynput precision) → Reasoning → Visual Perception (OCR) → Real-Time → Monet Visual Reasoning.

Lite ships the rule-based modes. Full adds the heavy AI and vision modes.

Confirmation before risky actions (toggle in Settings). A policy gate blocks destructive or sensitive actions.

Native control and coordinate clicks are off unless enabled. Press Esc anytime to stop instantly. Every run is logged.

Built for visible, user-directed actions — not unattended high-risk automation.

100+ movement patterns: absolute coordinates, named regions, percentages of the screen and relative moves.

Detailed pattern reference coming soon.

Monet-powered visual reasoning for complex interfaces (Full edition).

Setup guide coming soon.

Automates Excel workflows including formula entry and summaries, with overlay thought bubbles.

Full walkthrough coming soon.

Drive the Windows Calculator with natural language, with optional OCR-based visual verification.

Step-by-step guide coming soon.