phone-eye: Enabling AI Agents to Operate Real Android Phones via Visual and UI Tree Fusion

Phone-Eye is an open-source MCP plugin that uses ADB to allow AI agents to directly operate real Android devices without requiring additional app installation or root permissions. By fusing vision models with the UI tree, it provides core capabilities such as screenshot reading, taps and swipes, typing, and multi-device management. It supports automatic Wi-Fi reconnection and 24/7 unattended operation, making it suitable for scenarios such as mobile QA testing, form validation, and device monitoring. However, it currently only supports Android and cannot operate when the device is locked.