ESP32 iPhone Agent goes open-source, turning a low-cost ESP32 board into an AI controller for iPhone

ESP32 iPhone Agent goes open-source, turning a low-cost ESP32 board into an AI controller for iPhone

N
News Editor
2026-10-16 06:21:42
Developer Henry has open-sourced ESP32 iPhone Agent, a project that uses a XIAO ESP32S3 development board priced at roughly RMB 40 to 50 together with a mobile app to let AI operate a real iPhone. The setup can open apps, tap buttons, and swipe the screen without requiring a jailbreak or Apple XCTest testing privileges. Instead, the ESP32 creates its own Wi-Fi hotspot and acts as an AirPlay receiver. The iPhone mirrors its display to the board, which then sends compressed video back to an Agent app on the same phone over Bluetooth rather than decoding the stream itself. The app uses a vision model to read the interface, decides the next action, and sends commands back over Bluetooth so the ESP32 can emulate a USB mouse and keyboard. The system also calibrates the mouse pointer to reduce tap-position drift. The project runs on a standard ESP32-S3 and does not require dedicated video capture hardware such as NanoKVM Go, though it still has setup requirements and several technical limits.

Developer Henry has open-sourced ESP32 iPhone Agent, a project that uses a XIAO ESP32S3 development board costing about RMB 40 to 50 and a mobile app to let AI control a real iPhone. It can open other apps, tap buttons, and swipe across the screen.

No jailbreak and no XCTest access required

The setup does not require a jailbreak, and it does not rely on Apple XCTest testing permissions. Instead, the ESP32 creates its own Wi-Fi hotspot and serves as an AirPlay screen-mirroring receiver. The iPhone sends its display to the board, but the board does not decode the stream. It sends compressed video back over Bluetooth to the Agent app running on the same phone.

Vision recognition and input emulation

The Agent app calls a vision model to identify what is on screen and decide the next step. It then sends commands over Bluetooth, allowing the ESP32 to emulate a USB mouse and keyboard to operate the phone. To keep tap positions accurate, the system also calibrates the mouse pointer automatically and can correct the position based on the screen image when needed.

This design combines screen reception and input emulation on a standard ESP32-S3, without requiring dedicated video capture hardware such as NanoKVM Go. The Agent runs directly on the iPhone, while model inference depends on an external API or a self-hosted server. Because the phone’s Wi-Fi is occupied by the board’s hotspot, model access uses cellular data by default.

Current limitations

Initial firmware flashing, along with compiling and installing the iPhone app, still requires a Mac and Xcode. Bluetooth bandwidth limits the returned video feed to 480×1040 resolution at up to 5 frames per second, and each operation usually takes a few seconds.

At this stage, touch coordinates only work with portrait interfaces, keyboard input only supports ASCII characters, and screen mirroring stops once the phone is locked.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.
100

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.