AIshihzong/awesome-reverse-engineer-anything
Awesome REA (Reverse Engineer Anything): curated AI reverse engineering cases. What AI agents (REA, Craft Anything, Claude Code, Codex) reverse-engineered and rebuilt, open-source alternatives by app, and which ones really used REA. Unofficial, bilingual, every case checked.
What it solves
Reverse engineering closed software (native binaries, apps, firmware, websites) without source code is time-consuming and expert-only. Recently, AI agents have started doing this faster, but it's hard to know which tools actually work, what projects have been rebuilt, and whether REA (Reverse Engineer Anything) genuinely helps. This list centralizes those projects and clarifies what each case actually used.
How it works
It's an "awesome"-style curated list. Every project is tagged with its evidence source: REA (confirmed usage of morluto/rea), AI-RE (other AI tools like Claude Code + Ghidra), Rebuild (clean-room rewrite with no reverse-engineering evidence), or Claim? (unverified). The list is organized into official showcases, community cases (games, apps, hardware, security), and a wider wave section for projects that don't confirm REA. It also compiles controlled tests (rea-lab, leanrea) that compare REA against simpler approaches, and provides links to per-app indexes, a claim-check page, and a weekly-updates page. All entries include dates and point to source repos.
Who it's for
Anyone interested in AI-assisted reverse engineering: developers wanting to rebuild software they own, security researchers, hobbyists nostalgic for old games, and teams evaluating whether to adopt REA or similar tools. It's also useful for those building agents that inspect binaries and for people who want to verify claims about AI-driven reverse engineering.
Highlights
- Tracks dozens of projects where AI agents reverse-engineered and rebuilt software, from games like DX-Ball and GunZ to apps like a Bloomberg-style terminal and a macOS screen recorder.
- Covers REA itself (morluto/rea, 385k+ stars), an MCP server and CLI that orchestrates Ghidra, IDA, Hopper, and JADX for coding agents.
- Includes independent evidence: rea-lab's five studies (482 runs) found no significant solve-rate gain with REA but about twice the time/cost; leanrea's 36 runs showed a plain-CLI skill matching REA's accuracy (74/75 vs 75/75) with ~35% less cost.
- Provides official resources, showcases, and a community with submission guidelines and a Discord server.
- Includes defensive tools as well, like antirea and bla, which harden builds against AI reverse engineering.
- Offers structured docs: find projects by the app they replace, verify whether a case really used REA, and see new additions weekly.
Related
- Project
- Project
- Project
- Project