birkenpapier/ai-app-cloner

This project lets you recreate a mobile app from screenshots. You point an AI coding agent at a folder of screenshots taken from any phone app, and the agent figures out the screens, the navigation between them, and the visual design. It then writes a working React Native project that reproduces what you captured. The README shows six example clones built this way, including Discord, Todoist, and Google Keep, all running as real apps with working interactions rather than static images. The reason screenshots are the input comes down to what you can actually extract from a phone app. A website exposes its source code, styles, and assets through browser developer tools. A phone app gives you none of that. You cannot read the markup or copy the styling. You only have the pixels on the screen. Screenshots are the one thing you can capture from any app without special access, and the same folder of images works whether the app runs on iOS or Android. The core challenge is that an image contains no measurements. You cannot read a padding value off a screenshot. The tool solves this with a diff loop. It builds a screen, renders it, takes its own screenshot, compares that against your original, and fixes the differences. It repeats this cycle until the two images match. This loop replaces the convenience that website cloners get for free when they read computed styles from a web page. To use it, you need Node 18 or higher, a compatible AI coding agent, and a browser automation tool. You clone the repository, install dependencies, drop your screenshots into a designated folder, and run a command that triggers the cloning process. The output fills in the blank project you cloned, placing screens, components, and design details directly into the repository structure. There is no separate export step. The skill works across many AI coding agents including Claude Code, Cursor, GitHub Copilot, and others. Claude Code is the only one fully verified so far, with the rest using each agent's own configuration format generated from the same source files.