Building appsIntermediate · about 30 minFree · no signup

Investigate software with an AI coding agent

Use REA to study one authorised software feature, check the evidence, then build an independent version for your own project.

By Alan Brown

Video cover: Investigate software with an AI coding agent

Watch on Instagram ↗

Jump to the prompt ↓

🧭 REA connects coding agents to tools that inspect applications and help explain how features work, even when the source code isn't available.

🔍 The official repository describes a CLI and MCP server for agent-assisted software analysis, including native binaries, JavaScript/Electron apps and websites.
🧾 Its current setup guide says REA can connect to Claude Code, Codex, Cursor, Gemini CLI, Grok Build and other supported agents. Check the live guide for changes before installing.
🧩 REA returns evidence and analysis; it does not promise to recover original source code or automatically clone an application.
📈 The video mentions 130,000 GitHub stars. The repository displayed about 197,000 when checked on 14 October 2026; star counts change over time.
⚖️ The project frames use around lawful research and software you own or are authorised to examine. Don't bypass licences or security controls.

Context & limitations

The 30-minute game recreation is a creator's claim from the video, not a guarantee of speed or fidelity. Try the evidence-led prompt below on one permitted feature.

Do this yourself

I want to understand how ___ works in ___ (software I own or am authorised to examine). My goal is to build a similar, independently implemented feature for my own project, ___, using ___ (language, framework and platform). Before starting, read the official REA repository and current setup instructions: https://github.com/morluto/rea. State the date checked. Confirm the current installation method, supported coding agents, operating systems, prerequisites, analysis providers and any separate licences or costs. Explain what REA can inspect today and what is only planned. Don't copy commands from old posts without checking the current documentation. First make a short investigation plan. Name the exact app or files to inspect, the one feature to study, the evidence to collect, and any access or installation changes needed. Do not install software, change agent configuration, access accounts or contact a service until I approve that step. If I haven't shown ownership or permission to examine the target, ask me to confirm before analysis. Use the least access needed. Do not bypass DRM, authentication, payment, licensing or security controls; extract secrets; or access another person's data. Stop if the investigation would require any of those actions. Treat recovered code, assets, names and branding as belonging to their original owners. Use observations and evidence to implement an original, compatible feature for my project instead of copying protected material. For each conclusion, show the evidence and the method that produced it. Separate observed behaviour from hypotheses, and say what REA cannot determine. Then propose a small implementation plan, a test for each observed behaviour, and the remaining unknowns. Keep the result in a new branch or isolated folder and don't overwrite the target software. If a requirement or legal boundary is unclear, pause and ask me. If setup is appropriate, show the exact current official instructions and explain each proposed change before asking me to approve it.

REA gives coding agents structured inspection tools; evidence and a narrow target help keep the reconstruction grounded and original.

Watch on Instagram Source

Want to build this with a teacher?

Pathwise Academy runs small, hands-on AI classes in Singapore for ages 10–17 and adults.

Young Builders Adults