Reverse Engineer Anything: Clone Any App With ChatGPT or Claude
A free GitHub tool lets ChatGPT or Claude rebuild paid apps you already use. Here's how Reverse Engineer Anything works and how to try it.
What is Reverse Engineer Anything?
Reverse Engineer Anything is a free, open-source GitHub repository that you paste into a chat with ChatGPT or Claude to have the AI recreate an existing app or piece of software on your own machine. Instead of describing what you want and building it feature by feature, you point the AI at something that already exists (an app, a website, a tool you currently pay for) and it works backward from that finished product to rebuild its own version. There’s no subscription to the tool itself. You just need an AI plan capable of creating and managing files, and you need to run the desktop app version of ChatGPT or Claude, because the process depends on the AI being able to write files directly to your computer.
TL;DR
- Reverse Engineer Anything is a public GitHub repo you add to a chat conversation as a link, not a paid product or installed app.
- The method works by starting from the finished app and working backward to recreate its functionality, the opposite of the usual idea-to-plan-to-build workflow.
- You need the desktop app (ChatGPT or Claude), not the browser version, because the AI has to create and manipulate files on your local computer.
- Setting up a dedicated project folder with file access before you start the chat is what makes the whole thing work reliably.
- A demonstrated example recreated a screen-drawing utility (a paid tool called Screen Brush) in a couple of minutes of actual setup work, producing a standalone app called AIA Ink.
- Beyond desktop utilities, people have pointed the tool at web tools like Slido and even full creative suites, with one widely shared project recreating Adobe-style apps like Photoshop and Premiere Pro clones on a site called Artcraft.
- Because it’s just a prompt plus a public repo link, the barrier to trying it is low: no install, no technical background required beyond having a desktop AI app with file access.
● Reading src/server.ts
● Editing 3 files
+ import { limit } from 'rate'
- app.use(cors())
✓ Tests passed · ready for PR
How does Reverse Engineer Anything actually work?
The repo itself functions as a kind of instruction set or framework that tells the AI how to approach reverse engineering a target application. When you paste its GitHub URL into a chat, the AI reads those instructions and applies them to whatever you ask it to rebuild.
The underlying idea is a reversal of normal software development. Normally you start with an idea, turn it into a plan, then build and iterate. Reverse engineering flips that: you start with the finished thing and work backward to figure out what must have gone into making it, then reconstruct it. A useful comparison is cooking. Following a recipe means starting with ingredients and ending with a dish. Reverse engineering a dish means tasting the finished meal and working backward to guess the recipe. Reverse Engineer Anything applies that same backward logic to software, using the AI’s ability to infer structure, features, and likely code patterns from a description of what an existing app does.
What do you need to set this up?
The setup has a few concrete requirements:
- A desktop AI app. The browser version of ChatGPT or Claude won’t do here, because the process depends on creating actual files on your computer, not just generating text in a chat window. The desktop apps for ChatGPT, Claude, Claude Code, and tools like Codex can read and write to your file system.
- A higher-tier plan recommended. The process works best with more capable models, so switching to a higher model setting inside the app is worth doing before you start.
- A project with folder access. In ChatGPT, this means creating a new Project, giving it access to a specific folder on your computer (a fresh, empty folder is a clean way to start), and running the chat inside that project so the AI’s file operations are scoped to that location.
- The repo link itself. Once the project and folder are set up, you open a new chat inside that project and paste in the GitHub URL for Reverse Engineer Anything.
From there, the prompt can be as simple as asking the AI to “make me” a specific app that exists on your computer or that you use online. Running the model at a higher capability setting and giving it a short description of the functionality you want is often enough for a first working version.
Is Reverse Engineer Anything worth trying?
For anyone who pays for a utility app with more features than they use, or who wants a free, customizable version of a tool they rely on, it’s a low-risk experiment. There’s no cost for the repo itself, and the only real investment is the time spent setting up a project folder and writing a prompt. The demonstrated example, a screen-drawing utility recreated as “AIA Ink,” reportedly took just minutes of actual setup to produce a working, installed app, with follow-up messages used to adjust details like making the background transparent, something the original paid app didn’t offer.
- Propellant load✓ Go
- Range safety✓ Go
- Weather✓ Go
- Final pollT–00:42
The tradeoff is that you’re relying on the AI’s interpretation of how an app behaves rather than its actual source code, so results will vary by how well-known or well-documented the target app is, and by how complex its features are. A simple drawing overlay is a much easier target than a professional creative suite. That said, ambitious attempts have been made at recreating much larger software, including a project that reportedly rebuilt versions of Adobe tools like Photoshop and Premiere Pro and published them on a site called Artcraft, suggesting the ceiling for this technique is higher than it might first appear.
What can you actually rebuild with it?
The technique isn’t limited to desktop utilities. It’s been used to recreate web-based tools as well. One example involved rebuilding a copy of Slido, the audience Q&A and polling tool, closely enough that it replicated the core voting and question-submission experience. Because tools like ChatGPT can host simple web apps directly, the recreated version didn’t need to be deployed anywhere separately, it just worked inside the chat environment’s built-in hosting.
This points to a general pattern: the tool is well suited to apps whose core functionality is narrow and definable; drawing tools, polling tools, note utilities, simple productivity apps. The more a target app leans on a huge, deep feature set (like a full video editor or a word processor with decades of edge cases), the more iteration it will likely take to approach feature parity, even if the overall approach still works as a starting point.
Frequently Asked Questions
Do I need to pay for Reverse Engineer Anything?
No. It’s a free, public GitHub repository. The only costs involved are whatever subscription tier you already have for ChatGPT or Claude, ideally one of the higher plans for better results.
Can I use the regular web version of ChatGPT instead of the desktop app?
No, the process depends on the AI creating and editing files directly on your computer, which only the desktop apps (ChatGPT desktop, Claude desktop, Claude Code, or tools like Codex) can do.
Is this legal or is it just copying paid software?
The tool works by having an AI infer how an app’s functionality works and rebuild similar behavior from scratch, rather than copying proprietary source code. It’s best understood as recreating functionality you observe, similar to how developers have long built competing or compatible software by studying how existing tools behave.
What kinds of apps work best with this method?
Smaller, single-purpose tools with clearly defined features tend to work fastest and most reliably, things like screen-drawing utilities, polling tools, or simple productivity apps. Large, feature-dense software like professional creative suites is possible but will likely require much more iteration to approach the original.
What AI models have people used for this?
The method has been demonstrated with ChatGPT, and it’s also been noted to work well with Codex, Claude desktop, and Claude Code, since all of these can operate as desktop apps with local file access.
