Skip to content

Is the ultimate mobile Agent here? DIYgod open-sources Pinky, handing both iPhone and Android over to AI.

Nov 9, 11:48

Beating AI News Flash: RSSHub author DIYgod has open-sourced Pinky, enabling AI Agents such as Codex and Claude Code to directly operate real iPhone and Android phones. The desktop client supports Mac, Windows, and Linux. Once connected via MCP, the AI can view the screen, open apps, tap, swipe, and type text.

Many iPhone Use projects have been introduced before—some rely on XCTest/WDA testing interfaces, some simulate keyboards and mice via Bluetooth, and others require external hardware. Pinky integrates multiple technical approaches into a single desktop tool.

The Android side uses ADB + scrcpy, combined with an optional accessibility helper app to read UI controls. On Android 13 and above, a hidden virtual display can be created, allowing the AI to operate apps in the background while the user continues using the phone normally. This part is fairly similar to the previously introduced Mika.

For iPhone, different solutions are chosen depending on the OS version. On iOS 27 and above, it directly calls Apple's CoreDevice developer protocol to read the interface and send touch commands, without needing to install a test program—the technical approach is close to Poki. iOS 17.4 through 26 still relies on WDA, which requires signing and installation via Xcode.

Pinky also specifically addresses the problem of AI misoperations. It identifies sensitive buttons such as "Pay" and "Transfer," pausing before the actual tap and waiting for manual user approval. If there is no confirmation within 60 seconds, execution is refused; after approval, it re-checks the interface to prevent changes to the amount or buttons. Sending messages and deleting content can also be set to ask first.

This approval mechanism is written into the control program and does not rely on the model following prompts. Banks, wallets, and password managers are hidden from the Agent by default, and password and verification code input fields are prohibited from autofill. Users can take over the operation at any time, and every step is screen-recorded, with records saved on their own computer.

Project: https://github.com/DIYgod/pinky

Source