REA Lets AI Agents Figure Out How Apps Work, Even Without Source Code

REA takes the machine apart and works back to the plan.

You find a feature in an app that you wish you’d built. Maybe search feels instant. Maybe the clipboard keeps formatting when you paste into another program. You want to know how it works, and the source code isn’t available.

Normally, that’s where the investigation stops, because reverse engineering means specialized tools, unfamiliar code, and a lot of patience. REA (Reverse Engineer Anything) changes who can begin.

It’s an MIT-licensed, open-source project that has drawn a huge amount of attention on GitHub. It connects coding agents like Claude Code and Codex to tools that inspect software. You ask questions in plain English, your agent follows the clues through the app, and you get to examine the evidence behind its answers.

You don’t have to start with a decompiler, either. With JavaScript and Electron apps, you can begin with Node.js alone, or let your coding agent guide the investigation.

Here’s how to get there, what the evidence actually buys you, and where the AI can still get things wrong.

What REA actually is

REA is an MCP server. MCP (Model Context Protocol) is the plug standard that lets agents like Claude Code, Codex, Cursor, or Gemini CLI call outside tools. Once REA is…