Ever Wished You Could Pop the Hood on an App? REA Lets Your Agent Do It
You're using an app and you hit a feature that's exactly what you need in your own project—but there's no source code, no docs, and the behavior isn't obvious from the outside. That's usually where reverse engineering starts, and it's usually where most of us stop, because the tooling is scattered across a dozen programs and none of them talk to each other. REA is an attempt to fix that by putting reverse engineering tools behind a single MCP server your agent can drive.
What It Does
REA connects your agent to tools for inspecting native binaries, JavaScript and Electron apps, .NET assemblies, and websites. The pitch is straightforward: see a feature you like, ask your agent to investigate it, and get back an explanation of how it works—down to the binary level if that's where the answer lives. You can also call the same tools from your terminal if you'd rather work without an agent in the loop.
The project ships as an npm package (rea-agents) and runs analysis locally. That last part matters for anyone dealing with proprietary or sensitive binaries. Rather than handing you a confident-sounding answer with no backing, REA is built to include the evidence and the limitations behind each conclusion, so you can see what the analysis actually supports and where it stops.
Why It's Cool
-
One interface instead of a pile of tools. Native binaries, Electron apps, .NET assemblies, and websites all run through the same MCP setup. If you've ever stitched together a disassembler, a browser debugger, and a decompiler by hand, you know how much friction that removes.
-
Evidence over vibes. The README is explicit that results include the evidence and limitations behind each conclusion. For reverse engineering, that's the right design choice—an agent that says "this is probably how it works" is worse than useless, but one that shows you the disassembly it's basing that on is something you can actually check.
-
Local analysis. Everything runs on your machine. No uploading binaries to someone else's server, which is often a hard requirement in this space.
-
Terminal access, not just agent access. The MCP framing is the headline, but the same tools work from your shell. That means you're not locked into one workflow—use the agent when it helps, drive the tools directly when it doesn't.
-
It's positioned around a real task. The framing isn't "here's a collection of analysis utilities." It's "see a feature, understand it, build your own version." That's a concrete job, and the tooling is organized around it.
How to Try It
The quick start is a single command:
npx rea-agents setup
From there, you point your agent at the MCP server and start asking it to investigate things. If you prefer the terminal, the same tools are available directly.
A few practical notes before you install:
- Only get it from official sources. The README specifically warns about lookalike repository names and copied READMEs, so stick to morluto/rea, rea.tools, or the rea-agents npm package.
- Check the Node.js requirements before you start—the README links to the specifics under "What you can analyze."
- Browse the guides and showcases at rea.tools if you want to see what the workflow looks like on real targets before pointing it at your own.
The repository is at github.com/morluto/rea, and it's MIT licensed. There's also a Discord if you want to ask questions or see what other people are poking at.
Final Thoughts
REA is aimed at people who already know why they'd want to inspect a binary—security researchers, developers studying competitors' implementations, anyone doing interoperability work. If that's you, the combination of a unified MCP interface, local execution, and evidence-backed results is a sensible set of priorities. The honest caveat is that reverse engineering is hard, and an agent in front of the tools doesn't change that; it changes how much of the mechanical work you have to do yourself. That's a meaningful shift, and worth a look if the problem is one you run into.