storytold/craft-anything

Claims below include README evidence. A matching quote or worker agreement does not establish that a claim is correct. Category majority among 2 workers. Version 1.

Craft Anything is an open-source, pure Rust engine that studies software to understand its mechanics and uses AI agents to build new versions.

See how any website, app or game works, then have AI build your own version; an open-source, clean-room engine in pure Rust.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

It allows users to understand how any website, app, or game works and then have an AI coding agent build a new, clean-room version of it.

See how any website, app or game works, then have AI build your own version; an open-source, clean-room engine in pure Rust.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

It is a pure-Rust reverse-engineering engine with 320 tools that powers a desktop app, CLI, and MCP server, with parts running in the browser as WebAssembly.

A pure-Rust reverse-engineering engine with 320 tools (300 in the engine, the rest from the AI and setup layers) for native programs, .NET, Java, Android, Apple, Electron/JavaScript, websites, firmware, packet captures, databases and game files.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Studies websites, desktop apps, phone apps, games, and files using real inspection tools to provide evidence-backed answers.

Websites, desktop apps, phone apps, games and files: real inspection tools find out how they're built and what they do, with the evidence behind every answer.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Uses a coding agent to build a new project in the requested stack based on user instructions.

Say what you want ("Recreate this website", "Recreate this game's gameplay in Rust") and a coding agent builds a new project in the stack you ask for.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Compares the generated build with the original across multiple screen widths and test runs, refining until it matches.

Every build is compared with the original (pages at three screen widths, a game's own self-test, an app's runs) and refined until it's close.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Provides 320 tools available as both CLI commands and MCP tools, allowing agents to drive the engine without AI for inspection.

320 tools, each one a CLI command and an MCP tool, so any agent can drive the same engine the app uses. No AI is needed to inspect.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Operates in a default 'Clean room' mode that only uses visible elements like screenshots and layout, refusing to read internal code or files.

Clean room (the default): nothing inside the original is read, not its code, files or page source. Craft Anything works from what you can see: pictures you add, the website as a visitor sees it

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Offers a 'Deep' mode that reads code, files, and JavaScript using built-in engines and optionally Ghidra for private use.

Deep: every tool is used, for your private use. A program's code and files are read with the built-in engines (and Ghidra when it's installed), a website's JavaScript is read and its start page watched

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Supports Claude Code as the primary coding agent, running the user's unmodified program without handling their account credentials.

Claude Code is supported first: Craft Anything runs your own, unmodified claude program, which uses your own sign-in; Craft Anything never sees or handles your account.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Supports other ACP-speaking agents like Gemini CLI, Qwen Code, opencode, goose, Hermes, and Maki, with Maki being the only one verified live.

Other agents that speak ACP (Gemini CLI, Qwen Code, opencode, goose, Hermes, Maki) can be switched to it per backend; of those, only Maki has been run live.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Returns results as evidence records with confidence and coverage metrics, paging large outputs via core_evidence_browse.

Results are evidence records with their confidence, coverage and limitations; big ones come back as a summary first and are paged with core_evidence_browse.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Powers a desktop app using egui, a CLI, an MCP server, and runs parts of the engine in the browser via WebAssembly.

It powers the desktop app (egui), the CLI and the MCP server, and parts of it run in the browser as WebAssembly.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Inspects websites, desktop apps, phone apps, games, and files using real inspection tools to determine how they are built and what they do.

Websites, desktop apps, phone apps, games and files: real inspection tools find out how they're built and what they do, with the evidence behind every answer.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Provides 320 tools available as CLI commands and MCP tools, enabling any agent to drive the same reverse-engineering engine used by the application.

320 tools, each one a CLI command and an MCP tool, so any agent can drive the same engine the app uses.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Operates in a clean-room mode where no internal code or files are read, relying instead on visible screenshots, text, layout, and colors.

Nothing inside the original is read, not its code, files or page source. Craft Anything works from what you can see.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Offers a desktop app, CLI, and MCP server implementing a Study, Build, Check, and Refine workflow for websites, apps, and games.

The desktop app (Simple and Advanced), the CLI and the MCP server; the Study → Build → Check → Refine workflow for websites, apps and games.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Can generate pictures and 3D models as part of the build process, requiring an opt-in and a user-provided fal.ai key.

Generated pictures and 3D models (opt-in, your own fal.ai key).

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Returns results as evidence records containing confidence, coverage, and limitations, with large results summarized first and paginated for review.

Results are evidence records with their confidence, coverage and limitations; big ones come back as a summary first and are paged.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Explicitly does not bypass DRM, logins, or anti-cheat systems, focusing instead on recreating layout, style, and behavior from observable evidence.

It never bypasses DRM, logins or anti-cheat. Rebuilds recreate layout, style and behaviour from evidence.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

The coding agent acts on behalf of the user, meaning its ability to read, write, or run code depends entirely on the user's own system permissions.

A coding agent you run still acts as you, and what it can read, write or run depends on its own permissions.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Warning that installing project dependencies runs code on the user's computer, as packages may execute setup scripts during installation.

Installing dependencies runs code. When a builder installs a project's dependencies, the commands run on your computer, and the packages can run code while they install.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Requires Rust 1.97 or newer to build.

Requires Rust 1.97 or newer (a dependency needs it; rustup update).

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Building requires a coding agent signed in on the local computer.

Building needs a coding agent signed in on your computer (see AI); doctor says what's missing.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: The software is in early alpha status.

Craft Anything is in early alpha. What works today, what is partial and what's next:

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Must be used on software the user owns or is allowed to examine; it does not bypass DRM, logins, or anti-cheat.

Use Craft Anything on software you own or are allowed to examine. It never bypasses DRM, logins or anti-cheat.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Installing dependencies runs code on the user's computer, which can be a security risk if packages are untrusted.

When a builder installs a project's dependencies, the commands run on your computer, and the packages can run code while they install.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: The project is in early alpha status, meaning some features are partial or not yet fully implemented.

Craft Anything is in early alpha. What works today, what is partial and what's next.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Requires a coding agent installed on the user's computer, with Claude Code being the primary supported agent.

Building needs a coding agent on your computer. Claude Code is supported first.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Cannot bypass DRM, logins, or anti-cheat mechanisms, limiting its ability to recreate protected software.

It never bypasses DRM, logins or anti-cheat.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: Installing dependencies executes code on the user's machine, which may run arbitrary setup scripts from packages.

Installing dependencies runs code. When a builder installs a project's dependencies, the commands run on your computer, and the packages can run code while they install.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: The agent operates with the user's permissions, so it cannot access resources the user cannot access themselves.

A coding agent you run still acts as you, and what it can read, write or run depends on its own permissions.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

Limitation: The clean-room mode is not an OS sandbox and does not isolate the computer, relying instead on user-provided permissions.

Clean room is a rule about what Craft Anything offers and refuses. It is not an operating-system sandbox: it doesn't isolate your computer.

Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.

developer-tools · cli