REA turns coding agents into reverse-engineering assistants
REA (Reverse Engineer Anything) is an MIT-licensed MCP server and CLI that bridges coding agents like Claude Code, Cursor and Codex to tools such as Ghidra, IDA Pro and JADX, returning structured findings with evidence and confidence limits.
Mother of GOD REA stands for Reverse Engineer Anything. It’s a recent open-source (MIT-licensed) tool that turns AI coding agents (“bots” like Claude Code, Cursor, Codex, Gemini CLI, Grok Build, and others) into reverse-engineering assistants. What it does REA is an MCP (Model Context Protocol) server + CLI (rea-agents on npm). It does not replace decompilers. Instead, it acts as a bridge: It connects your coding agent to existing reverse-engineering tools (Ghidra, Hopper, IDA Pro for native binaries; JADX for Android; analysis for Electron/JS apps, .NET, websites, firmware, etc.). You ask the agent a plain-language question (e.g., “Understand how search works in this Notes app and show me the evidence,” or “Why does Windows Calculator do 200 + 10% = 220?”). The agent uses REA to inspect the binary/app, gets back structured findings (pseudocode, assembly, strings, function calls, module graphs, evidence + known limitations), then explains the behavior or even rebuilds a similar feature. Everything runs locally; results include evidence and confidence limits rather than pure guesses. Setup is typically just npx rea-agents setup (Node.js required). It registers with your agent and adds workflow instructions. It went viral in early–mid October 2026 (tens to hundreds of thousands of GitHub stars on morluto/rea) because it dramatically lowers the barrier to reverse-engineering work that used to require deep expertise in tools like Ghidra. W ho benefits Developers and engineers who want to understand how a closed-source feature, library, or app works so they can reimplement something similar, debug interoperability issues, or learn from it. Security researchers and reverse engineers who want to speed up analysis of binaries, malware (when authorized), APKs, firmware, etc., while keeping the agent’s reasoning tied to concrete evidence. Students and hobbyists learning reverse engineering, binary analysis, or how popular apps/games are built. Teams doing competitive analysis, protocol recovery, or reconstructing behavior for their own projects—again, only on software they are legally allowed to inspect. Official sources: GitHub morluto/rea and http://rea.tools. It’s moving very fast (multiple major versions in days), so check the latest release notes.