rea

The REA GitHub repository is one of the most remarkable open-source releases of 2026: an MCP server that puts professional reverse engineering tools in the hands of your AI coding agent. Instead of spending weeks inside a disassembler, you describe what you want to understand — an app’s behaviour, an undocumented API, a compiled binary — and REA’s agents probe, trace and reconstruct how the software actually works. With more than 398,000 stars and 99,000 forks, the REA GitHub project has become the reference point for agent-driven reverse engineering, and it is still moving fast. The official home of the project is the morluto/rea repository on GitHub, where you can browse the source code, releases and issue tracker.

What it is

REA (Reverse Engineer Anything), by developer morluto, is a single Model Context Protocol server written in TypeScript and released under the MIT licence. MCP is the standard that lets AI agents call external tools, and REA uses it to connect your agent — Claude Code, Cursor or another MCP client — to real reverse engineering engines instead of asking the model to guess from screenshots. You ask questions in plain language; REA routes the work to the right engine, runs the analysis locally on your machine, and reports back with evidence: the functions, strings and call paths behind each conclusion. Nothing about your target is uploaded to a third-party cloud, which is a decisive property when the software under the microscope is proprietary or sensitive.

Key features

Four kinds of target are supported out of the box, and each is treated as a first-class workflow:

• Native binaries — compiled executables analysed through professional engines: Ghidra, IDA Pro or Hopper, with optional assisted setup.
• JavaScript and Electron apps — static analysis of extracted app directories and ASAR archives, mapping modules, imports and process boundaries; no native engine required.
• .NET assemblies — inspection of managed code, types and metadata.
• Websites and web apps — analysis of behaviour and structure from the outside in.
• Evidence-first output — conclusions arrive with the artefacts that support them, so a human can audit the agent’s reasoning instead of trusting it.
• Local execution — targets and results stay on your machine; the official site publishes illustrated guides and case studies, and an active Discord supports users.

Who it is for

Security engineers auditing closed software, interoperability teams documenting an undocumented API, and maintainers resurrecting legacy systems are the obvious audiences. Agencies like ours also meet a quieter use case every month: a client arrives with a working product, no source code, and a vendor who has vanished. REA shortens the discovery phase of that rescue work from weeks to days. It is not a toy for casual curiosity, though — you need a legitimate right to analyse the target, and enough engineering judgement to sanity-check what the agent reports back.

Getting started

Clone the REA GitHub repository and follow the README’s MCP configuration for your agent of choice; the server is written in TypeScript, so a current Node.js runtime is the main prerequisite. Connect an analysis engine next — Ghidra is the capable free option, while IDA Pro and Hopper are supported if you already hold licences — and start with a small binary or app that you own and understand, so you can judge the quality of the output for yourself. Treat the first sessions as supervised analysis: ask narrow questions, inspect the evidence, then widen the scope gradually.

Our take

REA is the real thing, but it deserves a sober warning label. An agent that can reverse engineer anything can also be pointed at software you have no right to inspect, and the legal and ethical line sits with the operator, not the tool. Used legitimately, it is transformative: it converts scarce expert hours into directed questions and verifiable evidence. Used carelessly, its conclusions still need expert review — a model narrating a disassembly can be confidently wrong in ways that only show up later. We have covered this project in depth on the MTD blog — read our full deep-dive for the extended analysis, worked examples and responsible-use guidance. For teams doing security auditing, interop or legacy rescue, the REA GitHub repository belongs on your evaluation list.