What problem does it solve? It lets an AI agent directly operate an Android phone without root, a computer, or WiFi—locking the screen, killing apps, controlling media, reading the clipboard, and capturing screenshots for vision analysis through a local ADB loopback connection. ## Core Features & Use Cases - Device Control: Lock/wake the screen, force-stop apps, adjust volume, send media keys, and navigate with home/back keyevents via adb -s 127.0.0.1:5555 shell. - Screenshot + AI Vision: Capture the screen with screencap and analyze it using a vision script with Ollama, Qwen-VL, or GPT-4V providers. - Clipboard, URLs & Camera: Read/write the clipboard (with termux-api fallback), open URLs and deep links, and launch the camera. - Use Case: While you are away, the agent notices a misbehaving app, force-stops it, takes a screenshot to confirm the state, and sends you a notification—all over the local 127.0.0.1 ADB connection. ## Quick Start Ask the AI to take a screenshot of my phone and describe what is currently on the screen.