REA: reverse engineer anything with an AI agent
REA (Reverse Engineer Anything) is an open-source MCP server that gives an AI agent tools to inspect native binaries, Electron apps, .NET assemblies, Android APKs, firmware, model files and live websites, then explain how a feature works and help rebuild it. This guide covers what it does, how it works, setup, the full analysis matrix, showcases, and where it stops.
12 min read
Most reverse engineering starts the same way: you see a feature in an app you do not own, and you want to understand how it works, down to the binary. Traditionally that means downloading Ghidra or IDA, spending an afternoon orienting yourself, and hoping the function you care about is not inlined. REA (Reverse Engineer Anything) changes the entry point: you hand the target to an AI agent, and the agent drives the inspection for you. REA is an open-source MCP server plus CLI that exposes a large matrix of analysis tools, so the agent can inspect native binaries, JavaScript and Electron apps, .NET assemblies, Android APKs, firmware, model artifacts, live websites and more, then report findings with evidence and help you rebuild the feature.
What REA is
REA is a single MCP server (and a companion CLI, rea-agents on npm) that connects your AI agent to a toolkit for inspecting software without source code. Instead of you operating a decompiler by hand, the agent calls REA's tools, reads the results, asks follow-up questions, and either explains the behavior or writes and tests an implementation. Analysis runs locally on your machine; results include the evidence and the limitations behind each conclusion, so you can tell a solid finding from a guess.
- Open source under the MIT license, published as morluto/rea on GitHub and rea-agents on npm.
- 460k+ stars and 113k+ forks, with active development (release 6.9.x at the time of writing).
- Supports Claude Code, Codex, Cursor, Gemini CLI, Grok Build and other MCP-capable agents.
- Native analysis can reuse an existing Hopper, Ghidra, IDA or Kuna installation; REA can optionally install Hopper with approval.
- Static JavaScript and .NET analysis needs no native engine at all.
How REA works
Your agent talks to REA over MCP. REA inspects the target and returns structured findings with evidence. The agent uses those findings to ask the next question, explain the behavior, or write and test an implementation. The same workflows are available from the terminal through the CLI, so you can run one-off commands without an agent in the loop.
- Point the agent at a target: a feature you saw in an app, a binary, an APK, a website, a saved capture.
- The agent calls REA tools to inspect the target: decompile, extract strings, list imports, read metadata, capture runtime behavior.
- REA returns findings together with the evidence behind them and any limitations.
- The agent synthesizes an explanation or a reimplementation, and can run comparisons against the original.
See a feature you like. Understand how it works, down to the binary level.
Quick start
With Node.js and npm installed, run the setup command. It registers REA with your chosen agents and installs matching workflow instructions, with backups of existing configuration.
Set up your agent
- Run: npx rea-agents setup
- Choose your agents and review the proposed changes.
- Approve them, then restart your agent.
Static JavaScript and Electron analysis works immediately. Native analysis needs a decompiler provider configured first; setup can optionally install Hopper with approval. On Linux, Ghidra, JADX, firmware tools and static JavaScript/Electron analysis require a procps-compatible ps for process ownership checks.
Use the terminal directly
You can skip the agent and call REA from the command line. For example, to inspect an extracted JavaScript or Electron app directory:
npx -y rea-agents@latest analyze-javascript-application /absolute/path/to/app --json
The result includes modules, imports, Electron boundaries and their evidence. Replace the path with your target, such as D:/apps/example on Windows. To install the rea command for regular use: npm install --global rea-agents, then rea --help.
What you can analyze
REA's matrix is broad. It covers static inspection of many formats and live runtime capture, each with its own requirements. Highlights:
- Native binaries: pseudocode, assembly, strings, symbols, calls and references via Hopper, Ghidra, IDA or Kuna.
- Offline ELF layout: sections, segments, original symbols and relocations, static mitigation candidates.
- Go, Rust and Deno compile binaries: embedded module/version info and dependency edges.
- GPU code objects: CUDA/HIP targets, kernel parameters and retained PTX text.
- Model artifacts: GGUF and safetensors metadata, tensor declarations and derived totals.
- IL2CPP metadata, Godot PCK archives, Erlang BEAM, WASM, EVM bytecode.
- PE files: Authenticode certificates, imports, delay-load DLLs, hardening flags.
- JavaScript and Electron apps: modules, imports, source maps, routes, IPC, native add-ons.
- Android APKs: manifest, classes, decompiled methods, resources, signatures; plus live device inspection via adb.
- Firmware: regions and extraction via Binwalk/Unblob on Linux.
- Websites and saved network captures: page structure, scripts, requests, responses, exposed payloads.
- .NET assemblies: metadata, CIL instructions, native dependencies, build comparisons.
- SQLite databases, Apple IPS crashes, Linux kernel modules, recorded packets and logic signals.
- Process behavior: terminal output, interactions, exit status and filesystem observations.
Static inspection reads supplied files without running the application. Runtime capture runs or interacts with the selected target using your own permissions, and each runtime guide describes its effects.
Showcases
- DX-Ball: reconstruct a sound-pan calculation. A sound call is followed into its position-to-pan helper, instructions are inspected, and incomplete pseudocode becomes C. The reconstruction passes 3,205 original x86 cases and reproduces all 63 compiled function bytes.
- Notion: trace the Electron clipboard bridge. Find the renderer's clipboard API, follow it through preload and IPC into the main process, and inspect the rich clipboard format.
- TH04: recover a DOS bullet-ring calculation. Inspect the original PC-98 game's 16-bit instructions, recover fixed and aimed angle calculations, and compare reconstructed C++ with historical compiler output.
Build RL environments with REA
Beyond reading code, REA can record web and desktop workflows into reproducible environments. The recorded sessions can be turned into reinforcement-learning tasks, with the agent capturing the steps, state changes and outcomes that a model needs to learn from.
Important safety and trust notes
- Get REA only from morluto/rea, rea.tools, or the rea-agents npm package. A similar repository name or copied README does not establish that a download is official.
- Runtime capture executes or interacts with the target using your permissions; read each runtime guide before running.
- Native providers are caller-supplied: REA does not install or update Ghidra, IDA, Hopper, Kuna, objdump or Rizin for you.
- For large binaries, raise provider startup deadlines with REA_GHIDRA_STARTUP_TIMEOUT_MS or REA_HOPPER_STARTUP_TIMEOUT_MS.
- Keep the installation updated: rea update (or npx rea-agents@latest setup) pulls frequent bug fixes.
Limitations
- Native analysis needs a decompiler you already own; REA is the driver, not the engine.
- Results are evidence-based but not guarantees; complex, optimized or obfuscated code can defeat static analysis.
- Runtime capture has side effects on the target and on your machine, and requires the matching toolchain (adb, tshark, sigrok, etc.).
- Some features need specific host support (Linux x64 for several binary/runtime tools).
When to reach for REA
- You want to understand how a feature in a closed-source app works, with evidence you can cite.
- You need to port or reimplement a behavior and want a head start from the original.
- You are auditing a binary, APK or firmware for security and need structured findings.
- You are studying a compiled game or legacy program and want reconstructed C/C++.
- You want your AI agent to do the tedious decompilation work while you focus on decisions.
One MCP for reverse engineering across binaries, applications, and runtime behavior.
If you found this useful, browse the rest of the YesIntelligent articles for more practical guides on creator tools, repurposing video, and reading engagement honestly.
Related articles
Every free tool YesIntelligent offers, and when to use each
Nine small tools that cover the repetitive jobs of a creator's week: pulling the words out of a video, checking a post before it gets cut off, reading an engagement number honestly, turning a good line into an image, finding prospects, and rewriting stiff AI copy.
YouTube transcript generator: use cases and why ours
A YouTube transcript is the fastest way to turn a video you admire into text you can search, quote, and reuse. Here are the jobs it does best, and why ours stays out of your way.
Instagram Reel transcript generator: use cases and why ours
Reels are short, fast, and often uncaptioned. A transcript turns each one into text you can study in a minute, even when nothing is written on screen.