A $7 ESP32 Turns an iPhone Into the Eyes and Hands of an AI Agent

⚡ The gist in 5 seconds - An open-source tool called esp32-iphone-agent has been released — it lets any OpenAI-compatible vision model control a real iPhone. - The hardware is a $7 ESP32: eyes via AirPlay, hands via USB HID. - No developer account, Xcode, or jailbreak required. ### 🔍 What was found On GitHub, the repository onehumanbeing/esp32-iphone-agent publishes a project that turns a cheap ESP32 microcontroller into a bridge between an AI model and an iPhone. The setup works like this: the "eyes" receive the iPhone's screen image via AirPlay, while the "hands" emulate input via USB HID. This allows any vision-capable model compatible with the OpenAI API to see the screen and perform taps and swipes like a human. The author positions the solution as an alternative to official automation tools: the connection works without a paid developer account, without building in Xcode, and without a jailbreak. All you need is the iPhone itself, an ESP32 board, and an API key for a vision model. ### 💡 Why it matters This makes it cheaper and simpler to build AI agents that work with physical devices. Previously, such setups required either a Mac + Xcode, a jailbreak, or proprietary solutions. Now any enthusiast can assemble "hands and eyes" for an AI model from a $7 off-the-shelf component, opening the door to personal iPhone automation without the barriers of Apple's ecosystem. The project is brand new, so stability and compatibility with different ESP32 firmwares should still be verified manually.