How REA and Pi Coding Agent Simplify Reverse Engineering

Reverse engineering has always been tricky, demanding specific skills and tools. That's why Reverse Engineer Anything (REA) and Pi Coding Agent caught my attention. They're using AI to make things easier. Let me show you how they simplify digging into and recreating software functions.

Tools for the Job: REA and Pi Coding Agent

REA (Reverse Engineer Anything) isn't just another tool out there; it's an open-source kit fusing AI coding agents with classic reverse-engineering tools. With this combo, developers can wrap their heads around and replicate software features, even if the original source code is MIA. It's been generating buzz on GitHub from those looking into old software and unsupported gadgets.

What keeps REA ticking is its integration of AI agents like Pi, Claude Code, and Codex with tools such as Ghidra and Hopper, thanks to the Model Context Protocol (MCP). This protocol is crucial, ensuring AI agents mesh well with reverse-engineering tools for spot-on analysis.

Breaking Down the REA Mechanism

Let's take a closer look at the three main stages:

• Inspect: REA dives into binaries and software components to gather raw data. This means parsing the binary code and sorting out the application's core logic.
• Understand: The AI then digs into software functions, dependencies, and behaviors, fitting the pieces together to reveal how the app ticks.
• Recreate: With this understanding, it helps in coding and testing new functionalities.

This approach lets you work with hard facts rather than guesswork. Still, don't expect full source code retrieval, especially on platforms or applications where it might come up short.

Experimenting with Chrome's Dinosaur Game

So, how do REA and Pi break down and recreate Chrome's Dino Game?

1. Setting Up

First, prepare your Windows 11 environment:

• Chrome: You need this browser for the game.
• Node.js: Necessary for running REA and Pi.
• Pi Coding Agent and REA: These are your main players.
• AI Model: Use GLM-5.3 or similar, available via Z.AI or OpenRouter.

npm install -g --ignore-scripts @earendil-works/pi-coding-agent

Confirm your setup with these commands:

rea setup --client pi
rea doctor --client pi

2. Enabling Chrome Remote Debugging

Set up Chrome for remote debugging:

& "C:\Program Files\Google\Chrome\Application\chrome.exe" --remote-debugging-port=9222 --remote-debugging-address=127.0.0.1 --user-data-dir="C:\rea-chrome-profile"

You can check the remote connection like this:

(Invoke-RestMethod -Uri "http://127.0.0.1:9222/json/list") | Format-Table title, type, url -AutoSize

Setting up Chrome for remote debugging is crucial for REA to hook up with the software you want to analyze.

3. Exploring the Game's Mechanics

With everything in place, REA and Pi can unravel the Dinosaur Game's mechanics:

1. Start the Pi Coding Agent.
2. Use REA to identify and decode JavaScript functions like jumping, obstacles, scoring, and collisions.

Just keep in mind, REA doesn't work directly with chrome:// pages, so it leverages JavaScript from local Chrome resources, showing the need for adaptation in unsupportive environments.

4. Crafting a Game Clone

Once the game mechanics are understood, Pi steps in to build a new game:

Ask Pi to develop a fresh endless runner with the same mechanics, but a new look:

Prompt: Use what we learned to build a simple browser-based endless runner.

This process transforms the classic Dinosaur Game into Neon Rooftops, featuring an orange cat leaping across a cityscape—same fun, different vibe.

5. Creating a Standalone App

For easier access, you can convert the browser game to a standalone app with Electron:

Direct Pi:

Prompt: Convert our game into a standalone Windows application.

The output is a game that runs offline without needing web servers or browsers. Electron packages ensure everything runs smoothly.

What I'd Consider

While REA and Pi are undoubtedly useful, there are some things I'd keep in mind:

• Legal Issues: Reverse engineering is powerful, but it's essential to stay ethical and legal. Focus on learning and improving features rather than copying outright.
• Platform Limits: Some platforms or apps can resist analysis or replication, particularly those with tough DRM or obfuscation, which could limit what REA can achieve.

In the grand scheme of things, REA and Pi open up new ways to explore legacy software, expanding your toolkit for working with existing applications.