Adding REA to Hermes Agent and OpenClaw

REA, short for Reverse Engineer Anything, is an open source tool that lets an AI agent take apart software that is already built and shipped, and explain how it works. The project had 255,419 stars on GitHub when OneClickClaw read the count on the 15th of October 2026. Version 6.5 was published on the 14th of October 2026. It plugs into Hermes Agent through its own setup step, and into OpenClaw through OpenClaw's settings for tools and skills.

What REA does

The project describes itself on GitHub as a way to "reverse engineer anything with agents, from app behavior down to native binaries". In plain words, reverse engineering means working backwards from a finished program to understand what is inside it.

REA is an MCP server. MCP is a standard plug that gives an AI agent extra tools. REA also ships with a skill, a written instruction file that tells the agent how to use those tools.

The README lists a long range of targets:

• desktop programs, read through the analysis engines Hopper, Ghidra or IDA
• programs written in Go and Rust, and WebAssembly code
• JavaScript and Electron desktop apps, websites and recorded web traffic
• Android apps and devices, Flutter apps and .NET programs
• firmware, smart contract code, SQLite databases, network captures and crash recordings

The README names the uses it is meant for: research, education, making two programs work together, preserving old software, and studying software you own or are allowed to examine. The project site shows worked examples on Windows Calculator, the Chrome dinosaur game, an Android authenticator app and the Notion clipboard.

REA is released under the open source MIT licence. It needs Node.js 22 or newer, the program that runs JavaScript tools on a computer. The heavy analysis engines, such as Ghidra or IDA, are installed separately. The setup step can install Hopper, and only with the user's approval. JavaScript and Electron apps need no extra engine at all.

What people did with it in its first busy week

The repository was created on the 14th of April 2026 and counted 64,181 forks, or personal copies, by the 15th of October 2026. Version 6.0 came out on the 8th of October and version 6.5 six days later. Users posted their results in the days that followed.

On the 10th of October 2026, a user ran an outside in leak check on their own live web shop, built with Next.js and Stripe.

On the 12th of October 2026, systemdesign.academy tested REA on the Electron API Demos app. It found 30 of 30 code locations correctly. It found 18 message channels, where a plain text search found 17. The same test listed what REA missed: all 8 reply senders and one security setting, and it counted 1 window where the app has 8.

The same day, a user in the project's Show and Tell discussions built a desktop app in the style of QuickBooks. Another user checked REA's reading of a stripped bzip2 program against its source code and documentation. On the 14th of October 2026, that user ran REA with a small local model, with a memory of 32 thousand tokens and 10 tools, and passed 3 of 3 tasks.

On the 15th of October 2026, the project's issues page listed open bug reports, among them Gemini clients refusing the whole tool list.

How to add REA to Hermes Agent

REA's get started page offers one setup step that adds the REA connection and the matching skill to the agent, and keeps a backup of the existing settings first. Hermes is one of the agents that page names, along with Claude Code, Codex, Cursor, Gemini CLI, Windsurf and OpenCode.

Hermes Agent keeps its MCP servers in its own configuration file, under a section for MCP servers, and has its own command to add one. Skills live in the Hermes skills folder and install with the Hermes skills command. So a Hermes user can either run REA's setup and let it fill in both parts, or add the REA server and the REA skill through those two Hermes settings by hand.

How to add REA to OpenClaw

OpenClaw users add REA through OpenClaw's own settings, in two parts.

First, the tool connection. OpenClaw keeps MCP servers in its configuration, under a section for MCP servers, each one listed as a command that runs on the same machine. OpenClaw also has its own command to add one. REA goes in as one entry there.

Second, the skill. In OpenClaw a skill is a folder holding an instruction file, kept in the skills folder of the workspace. OpenClaw has a skills install command for this. The REA skill goes in as one folder.

With both parts in place, the OpenClaw agent sees the REA tools and the instructions for using them.

The project's own line on use

The README sets the terms in one sentence: "REA is for lawful research, education, interoperability, software preservation, and analysis of software you own or are authorized to examine". The project's FAQ adds that "REA analyzes targets on your machine".

People who run OpenClaw or Hermes Agent on OneClickClaw add MCP servers and skills in their own agent's settings. Plans start at Starter, EUR 14.99 per month, with a seven day free trial.