ToutCas

Nothing stored. Nothing shared. A Burn-After-Use AI assistant.

It is an AI assistant designed with a Burn-After-Use interaction model and, zero remote data persistence, and a secure multi-tenant architecture, ensuring that user data is never stored or retained on remote servers.

Features

• ✅ An agentic AI web browser, inspired by ChatGPT Atlas
• ✅ Burn-After-Use (BAU) mechanism with a non-custodial design, ensuring no persistent storage of user data during AI interactions
• ✅ Cross-platform support: macOS, Windows, and Linux, (iOS and Android versions are under development)
• ✅ Built-in support for multiple LLMs, including gpt-oss-120b, DeepSeek V3.2 Exp, and Mistral-7B-Instruct-v0.2
• ✅ Native web page analysis and PDF understanding powered by Retrieval-Augmented Generation (RAG) and embedding-based retrieval
• ✅ Research-inspired design, informed by this dissertation arXiv:2601.06627

Screenshots

Environment

Python + AI

[✓] Python (version 3.9.6)
[✓] Flask (version 3.1.2)
[✓] transformers (version 4.57.3)
[✓] torch (version 2.9.1)
[✓] tokenizers (version 0.22.1)
[✓] scikit-learn (version 1.8.0)
[✓] numpy (version 2.3.5)
[✓] huggingface-hub (version 0.36.0)
[✓] scipy (version 1.16.3)
[✓] sentence-transformers (version 5.2.0)

Client side

[✓] Flutter (Channel stable, macOS 15.6.1 24G90)
    • Flutter version 3.35.1
    • Dart version 3.9.0

[✓] Android toolchain (Android SDK version 35.0.0) 
    • Emulator version 34.1.19.0  
    • Platform android-35, build-tools 35.0.0
    • All Android licenses accepted.

[✓] Xcode (Xcode 26.1.1) 
    • Build 17B100
    • CocoaPods version 1.16.2

[✓] Chrome - develop for the web
    • Chrome at /Applications/Google Chrome.app/Contents/MacOS/Google Chrome

[✓] Android Studio (version 2024.2)
    • Android Studio at /Applications/Android Studio.app/Contents
    • Java version OpenJDK Runtime Environment (build 21.0.3+-79915917-b509.11)

[✓] VS Code (version 1.106.2)
    • VS Code at /Applications/Visual Studio Code.app/Contents
    • Flutter extension version 3.122.0

Tip

Use it at your own risk

Theme of the Thesis

Tip

Burn-After-Use for Preventing Data Leakage through a Secure Multi-Tenant Architecture in Enterprise LLM

Citation

If you use ToutCas in your research or academic work, please cite:

@misc{ToutCas2025,
  title        = {ToutCas: An AI Personal Browser with a Burn-After-Use Interaction Model and Non-Custodial Data Handling},
  author       = {Zhang, Victor},
  year         = {2025},
  howpublished = {\url{https://github.com/VictorZhang2014/toutcas}},
  note         = {arXiv:2601.06627}
}