CLI tool
morluto/rea avatar
morluto/rea

REA (morluto/rea): an MCP server that lets coding agents drive Hopper, Ghidra, IDA and Kuna

Reverse engineer anything with agents, from app behavior down to native binaries.

471,216 stars115,069 forksTypeScriptMIT

At a glance

What is it?
REA packages reverse engineering tools for native binaries, Electron apps, .NET assemblies and websites behind one MCP server and CLI. The JavaScript path works on its own; the native path depends on a decompiler you already have, a recent Node release and, on Linux, a procps-compatible ps.
Who is it for?
Use REA if you already run Hopper, Ghidra, IDA or Kuna and want a coding agent to drive them with evidence attached, or if you want an agent to inspect Electron and JavaScript apps without any decompiler. Skip it if you expected a self-contained native decompiler in the npm package.
Can I use it commercially?
Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository received new commits within the last day.
What is it written in?
Mainly TypeScript, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on November 20, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • morluto/rea471,216 ★ · +470,815
Star history
Repository2026-09-122026-11-20
morluto/rea401471.2k

470,815 Stars in the last 69 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

One MCP server in front of Hopper, Ghidra, IDA and Kuna

REA, published on npm as rea-agents, is a TypeScript project under the MIT license. It exposes reverse engineering tools to coding agents through the Model Context Protocol and exposes the same tools to a shell through a CLI. Its targets are native binaries, JavaScript and Electron apps, .NET assemblies and websites.

REA does not bring its own native disassembler. For native work it drives an existing Hopper, Ghidra, IDA or Kuna installation, and the npm package carries the glue for that: a Hopper bridge in Python, a Ghidra bridge written in Java, a Cutter bridge script, Swift sources for native UI inspection and process tokens, and an LLDB tracer script. Analysis runs on your own machine, and each conclusion comes back with the evidence and the limitations behind it. The intended loop is concrete: point your agent at a feature in an app whose source you do not have, ask how it works, see the evidence, and have the agent build a similar feature in your project.

npx rea-agents setup edits your agent configuration, with backups

With Node.js and npm installed, the entry point is one command:

bash
npx rea-agents setup

Setup asks which agents to configure, shows the proposed changes, and writes only after you approve. It registers REA's MCP server, installs matching workflow instructions, and keeps backups of the configuration it touches. Restart the agent afterward. Named targets are Claude Code, Codex, Cursor, Gemini CLI and Grok Build, with more listed in docs/installation.md, which also covers manual MCP registration and plugin installation for Claude Code and Codex.

Two details have consequences later. Setup records the selected runtime and the REA entry script directly, so if you acquired REA through a package runner, deleting that runner's cache entry leaves the registration pointing at nothing. And package acquisition should run from a trusted directory outside the target project, not from inside the app folder you are about to inspect.

Native analysis stays unavailable until you configure a provider

Installing the npm package does not give you binary analysis. Before any native command runs, you have to configure a provider, which means having Hopper, Ghidra, IDA or Kuna installed and selected. Setup can install Hopper for you, and only with your approval. For the other three you bring your own installation, and any licence or platform constraint of that tool comes with it.

Provider selection, the native commands, snapshots and scripting are covered in docs/cli.md, the CLI and Evidence guide. The consequence for readers is simple: if you do not already run one of these four engines, the native half of REA is not something you get for free. The depth of what an agent can report about a binary is bounded by the decompiler underneath. REA is the layer that lets an agent operate those tools and collect evidence from them, not a substitute for them.

On Linux, a missing procps ps blocks most analyzers

REA performs process ownership checks before it launches several tools, and on Linux those checks need a procps-compatible ps on REA's PATH. The affected work is broad: Ghidra launches, JADX launches, firmware tool launches, and static JavaScript and Electron analysis.

That last item is the one that catches people. The JavaScript path is the part of REA that needs no native engine and is otherwise the easiest place to start, yet on a Linux host without a compatible ps it fails the same way the Ghidra path does. Minimal container images and systems that ship a different ps implementation are where this shows up. The recovery steps live in docs/installation.md under the heading "Linux owned-process prerequisite". The repository includes an examples/github-actions/ directory for running REA in CI; check that your runner image provides a procps-compatible ps before relying on it.

analyze-javascript-application reads an Electron app without a decompiler

The terminal path for JavaScript targets takes an extracted JavaScript or Electron app directory, or an ASAR archive:

bash
npx -y rea-agents@latest analyze-javascript-application /absolute/path/to/app --json

The result lists modules, imports and Electron boundaries, each with its evidence. The path must be absolute; on Windows the form is "D:/apps/example". The --json flag makes the output machine readable, which is what an agent or a script consumes.

The npx form with @latest resolves the newest published version every time it runs. For regular use, install the CLI globally instead:

bash
npm install --global rea-agents
rea --help

The package defines two binaries, rea and rea-agents, both pointing at scripts/rea.mjs, so either name works after a global install. JavaScript analysis and web bundle analysis run in separate worker processes inside the package.

Compact input schemas and full output schemas by default

REA trims what it advertises to the agent, in one direction only. Input schemas are compact by default to reduce context overhead. Setting REA_MCP_INPUT_SCHEMA_PROFILE=full in your MCP registration restores the complete field guidance and nested validation details.

Output schemas go the other way: they are complete by default. If your client caps the size of the complete tools/list frame, REA_MCP_OUTPUT_SCHEMA_PROFILE=compact bounds each advertised output schema, and off omits output schemas entirely.

None of these settings remove tools. Both input profiles keep the full tool set and canonical server validation, and both output settings keep server-side validation and complete structured results. In practice, the full input profile is the knob to turn when an agent needs more guidance about arguments, and the output profile is the knob to turn when a client refuses to load the server's tool list. Details are in the MCP contracts guide under "Compact input schema profile" and "Output schemas".

Three releases in eight days and a narrow Node engine range

The project moves fast. rea-agents 6.6.0 and 6.7.0 both shipped on 18 October 2026, 6.8.0 followed on 25 October, and the last push to the repository was on 2 November 2026. The project states that REA changes quickly, that new releases include frequent bug fixes, and that users should keep installations current:

bash
rea update

rea --update is an alias for the same verified workflow. It applies to CLIs installed with npm install --global or bun add --global. Bun global updates need Bun 1.2.17 or later, so upgrade Bun by hand first if yours is older. The updater retains at most 4 MiB of stdout and stderr combined and rejects oversized metadata replies.

The other constraint is Node. package.json declares node ^22.23.4 || ^24.22.1 || >=26.12.1. A Node 22 or 24 install below those patch releases falls outside the range, and odd-numbered lines such as 23 and 25 do not match. Machines pinned to older Node need an upgrade before REA will install cleanly.

Only three download sources count as official

The project carries an explicit warning about provenance: get REA from the morluto/rea repository, from rea.tools, or from the rea-agents npm package, and follow installation and release links only from those. A similar repository name or a copied README does not establish that a download is official.

For a tool that you register with a coding agent and that launches local analysis processes, this is the check to make before anything else. Its MCP registry name is io.github.morluto/rea, which gives you a second identifier to match. The repository also includes DISCLAIMER.md and SECURITY.md at the top level, and the README is available in more than twenty languages. Guides and case studies are published at rea.tools, and community questions go through the project's Discord server.

Editorial conclusion

Use REA if you already run Hopper, Ghidra, IDA or Kuna and want a coding agent to drive them with evidence attached, or if you want an agent to inspect Electron and JavaScript apps without any decompiler. Skip it if you expected a self-contained native decompiler in the npm package. Before installing, confirm your Node version falls within ^22.23.4, ^24.22.1 or >=26.12.1, that Linux hosts have a procps-compatible ps on PATH, and that you are pulling from morluto/rea, rea.tools or the rea-agents npm package.

Frequently asked questions

How do I use morluto/rea?

With Node.js and npm installed, run npx rea-agents setup, approve the proposed changes for your agent, and restart the agent. Then ask the agent to investigate a feature in a target app, or run commands such as analyze-javascript-application from the terminal.

Does REA need Ghidra, IDA or Hopper installed?

Only for native analysis, which uses an existing Hopper, Ghidra, IDA or Kuna installation; setup can optionally install Hopper with your approval. Static JavaScript and Electron analysis needs no native analysis engine.

Which agents can REA setup configure?

Setup supports Claude Code, Codex, Cursor, Gemini CLI and Grok Build, with other agents listed in docs/installation.md. That guide also covers manual MCP registration.

Official sources

  1. Issues
  2. License: MIT
  3. morluto/rea on GitHub
  4. README
  5. Releases
Community

Where developers are discussing it

Posts on Hacker News, dev.to and Lobsters that link to this repository, found by our scan of those communities. 2 in the last 30 days; newest first.

Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/morluto-rea.svg)](https://hysenlabs.com/projects/morluto-rea)

Weekly Top 10 badge

markdown
[![Hysen Labs](https://hysenlabs.com/badge/morluto-rea.svg?metric=rank)](https://hysenlabs.com/projects/morluto-rea)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/morluto-rea.svg)](https://hysenlabs.com/projects/morluto-rea)

Was this useful?