ESP32 iPhone Agent: $7 AI control, no jailbreak needed

Someone brilliant figured out how to let an AI model use a real iPhone. All it takes is a $7 ESP32 microcontroller. This workaround bypasses Apple's usual restrictions, making your iPhone an AI-controllable tool.

Developer onehumanbeing published the esp32-iphone-agent project on GitHub. Forget expensive Macs or complicated jailbreaking procedures. This system lets any OpenAI-compatible vision model "see" the iPhone's screen via AirPlay. Then, the ESP32 "clicks" and "types" by simulating a USB keyboard or mouse. Who knew a $7 chip could pull off such a feat?

The trick is separating the AI's "hands" and "eyes." The iPhone streams its screen via AirPlay to a Raspberry Pi or PC, which then feeds it to the vision model. The AI decides what to do, sending commands to the ESP32.

The ESP32 simulates a USB Human Interface Device (HID), allowing it to send input like taps and typing to the iPhone. It's like having a ghost operate your phone. The AI sees what you see, figures out its next move, and then the ESP32 physically executes the commands. All this happens without ever touching the iPhone's internal software. Pretty clever, right?