Source · repo
REA: reverse engineering with agents (MCP server and CLI)
Summary
REA is an MIT-licensed MCP server and CLI that lets coding agents inspect software
without its source: native binaries, JavaScript and Electron apps, .NET assemblies,
Android APKs, websites, firmware, and packet captures. The agent explains how a feature
works and shows the evidence. date_published is the date the README was read; the repo
has no releases.
Key points
- Install:
npx rea-agents setupproposes config changes for your agents (review them, then restart the agent), ornpm install --global rea-agentsandrea --help. Needs Node.js 22.23.4+, 24.22.1+, or 26.12.1+. - Agents: Claude Code, Codex, Cursor, Gemini CLI, Grok Build, and any client via manual MCP registration.
- Engines: static JavaScript analysis works on its own; native analysis drives an existing disassembler (Ghidra, Hopper, IDA, or Kuna) that you must obtain legally.
- Good for: understanding an undocumented dependency or binary, debugging a closed third-party app's behaviour, malware triage, and checking what an Electron app does.
- Caution: the README warns that look-alike repos copy its name; install only from
the repo, rea.tools, or the
rea-agentsnpm package. The star count (468k on 2026-11-19, +15k in a week) is unusually high, but the repo shows about 8,450 commits and 670 open issues, so it is actively used.
Why it matters
It turns a coding agent into a reverse-engineering assistant using tools many security engineers already have. Only analyse software you are allowed to.