Pickcraft-anything: A Rust engine that analyzes and recreates websites and games with AI

storytold/craft-anything

·2026.11.02 12:42

Craft Anything is an open-source tool that analyzes the internal structure of various software, including websites, desktop apps, mobile apps, and games, to generate new projects based on that analysis. It features a reverse engineering engine written in pure Rust that determines how the target software works, after which a coding agent builds a new project using the stack specified by the user.

1 / 4

Beyond simple analysis, the process includes verifying the generated output by comparing it with the original. Websites are checked for similarity to the original across three screen widths, while games undergo their own tests to verify similarity, looping through corrections for any discrepancies. This process supports a clean-room mode that works solely with external observation data without reading the original code directly, and a deep mode that analyzes internal code using tools such as Ghidra.

All 314 tools are exposed as CLI commands and MCP tools, allowing agents to run the same engine without an app. It prioritizes support for Claude Code and utilizes agents that users log in to directly, so account information is not passed to the tool side. It handles a wide range of target formats, including native programs, .NET, Java, Android, Electron, firmware, packet captures, databases, and game files.

Currently in early alpha, the primary features are the recreation workflows for websites and games. It supports game recreation using Unity, Unreal, and Godot engines, as well as generated images and 3D models. It is suitable for developers who want to understand software structure and build alternatives with similar functionality, or analyze the operating principles of existing projects.

GitHub
GitHub repository

storytold/craft-anything

Craft Anything: understand how any software works, and craft your own.

Rust

This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.