MarkClean: Claude Watermark Remover

MarkClean finds and removes hidden watermarks that Claude (recent update), ChatGPT, and other AI tools quietly insert into generated text - invisible characters you can't see, but that can be used to trace where a piece of text came from.

It works in two steps:

1. Layer A: Invisible character removal. Scans your text for known invisible/hidden characters and strips them out. This part is instant, exact, and runs on your device.
2. Layer B: LLM rewrite (optional). Sends the cleaned text to a local AI model to rewrite it in different words, as extra protection against watermarking techniques that don't use hidden characters at all. This step is optional and off by default you turn it on only if you want it.

Everything runs on your own computer. Nothing is uploaded to a company server.

---

Fastest way to run it: Claude Code

If you have Claude Code installed, you can set this whole thing up - download the app, install the local AI model it needs, and start it, by just pasting one message into Claude Code:

Clone https://github.com/Rjgyana/MarkClean-ClaudeWatermarkRemover.git, install its dependencies, install Ollama if it isn't already installed, pull the llama3.2 model with Ollama, then start the app with npm run dev and open it in the browser for me.

Paste that into Claude Code and it will handle every step — cloning the project, installing everything it needs, downloading the local AI model (a few GB, one-time), and opening the app for you. No prior coding experience needed.

---

Running it manually

If you'd rather do it yourself step by step:

1. Install Node.js (the runtime this app needs) if you don't have it.
2. Get the code:

   git clone https://github.com/Rjgyana/MarkClean-ClaudeWatermarkRemover.git
   cd MarkClean-ClaudeWatermarkRemover
   npm install

3. (Optional, only needed for Layer B) Install Ollama — this lets you run an AI model locally for the rewrite step, for free, with no API key. After installing it, download a model:

   ollama pull llama3.2

4. Start the app:

   npm run dev

5. Open http://localhost:3000 in your browser.

You don't need Ollama to use Layer A (the invisible-character removal) — that part works immediately with no setup. Ollama is only needed if you turn on the LLM Rewrite toggle.

---

License

This project is free to use for personal, educational, and research purposes, including modifying it for yourself. It may not be resold, rebranded/white-labeled as someone else's product, or hosted as a paid public service. See LICENSE for the full terms.