Reverse Engineer Anything, or REA, is a free, MIT-licensed open-source tool that connects Claude Code and other agents to reverse-engineering tools. You install it with one command, npx rea-agents setup, point Claude Code at an app on your own machine, ask how a feature works, and it returns the code paths it traced, the evidence behind each step and the parts it could not confirm. Two things the viral posts skip: it is not any mobile app, because the README lists desktop, Electron, web and Android targets and no iOS target, and the project's own disclaimer limits use to software you own or are authorized to examine. Used that way, it is the fastest route I know from a feature you admire to a pattern you can build yourself. The presenter in the reel is an AI avatar of Hammad; the words and facts are his.
What REA does inside Claude Code
Setup adds REA's MCP server and workflow instructions to Claude Code. After that, Claude can drive the reverse-engineering tools already on your machine and report back like an investigator: here is the feature, here is each step from the user's click to the final output, here is the evidence for each step, and here is what I could not confirm. That last part is what makes it useful rather than impressive. A tool that states its unknowns is one you can plan a build on; a tool that sounds confident about everything is one you have to re-check line by line.
What REA can inspect, and what it cannot
The README's support table covers native desktop binaries (with Hopper, Ghidra or IDA installed), JavaScript and Electron apps, websites (with a Chrome-family browser), Android APKs (with JADX and a JDK), .NET assemblies and more. There is no iOS app support listed, so the claim that it takes apart any mobile app is not accurate. The external tools are your responsibility: REA orchestrates Ghidra or JADX, it does not replace them, so an Android investigation needs JADX and a JDK on the machine before Claude can use them.
The rule: only software you own or are authorized to examine
This is the project's own position, written in its DISCLAIMER, and it is the rule I follow. Point REA at your own product, a client's product with permission, or software whose licence allows inspection. Use what you learn to understand patterns and build your own version, not to copy someone else's code. My three prompts are built around that boundary: the first asks for evidence, the second asks for the pattern in plain language and explicitly forbids reproducing the original code, and the third asks for an original implementation kept independent from the inspected app.
REA setup for Claude Code in about 5 minutes
Install Node.js 22.19 or newer; the README lists 22.x at 22.19 or above, 24.x at 24.11 or above, or 26 and up. In a terminal run npx rea-agents setup, choose Claude Code, review the proposed changes and approve them. Restart Claude Code and confirm the REA tools are available. One warning from the README that matters: install only from the official repository, morluto/rea on GitHub, or from rea.tools, because look-alike repositories exist. The npm package is rea-agents, MIT-licensed, at version 6.5.0 when I checked.
ONE TACTIC A WEEK
Prompt 1: map the feature with evidence
Use REA to investigate [app path or URL]. I own this app or am authorized to examine it. Explain how the [feature name] feature works from the user's click to the final output. For every step, show the evidence REA found and list anything you could not confirm. Replace the two placeholders and keep the authorization sentence in; it sets the frame for everything Claude does next.
Prompt 2: extract the pattern, not the code
From that investigation, describe the underlying pattern in plain language: data flow, main components, and the key decisions. Do not reproduce the original code. Then list the 3 parts that would be hardest to build and why. This is the step people skip, and it is the one that turns an investigation into something you can build from. A pattern description is portable across stacks; a copied function is not, and it is also not yours.
Prompt 3: build your own version on your stack
Design an original implementation of this pattern for my project. My stack: [framework, language, database]. Give me a file-by-file plan, then implement the first file with tests. Keep it independent from the inspected app's code. Fill in your stack, run it, and you have the start of your own feature with tests, built from an understanding of how the original works rather than from its source.
Get the setup guide and the 3 prompts
The free guide at withhammad.com/resources/reverse-engineer-features-with-claude-code has the setup steps, the support table, the responsible-use rule and all three prompts ready to copy. On Instagram, comment BLUEPRINT on the reel and the guide link arrives by DM (follow @withhammad.ai first so the DM reaches you).
Related reading
If this is useful, see the 27 free Claude skills for LinkedIn and how to scan Claude skills with Skillspector before installing them.