The memory layer for every AI model. Learn more » · Issues
0428.mp4
Havril is a model-agnostic memory service for AI models. It lets Claude, ChatGPT, Gemini, Mistral, and any other model remember you persistently across conversations — without changing how you use them.
You keep chatting on Claude.ai, ChatGPT.com, or wherever you already are. Havril works silently in the background: after each conversation, its Memory Engine distills what matters and stores it. Next time — on any model — that context is already there.
The intelligence lives inside Havril, not the models. Models are just delivery pipes. They submit raw conversations and receive distilled memories back. Havril decides what is worth keeping, how important it is, and when it conflicts with something older.
Every AI model starts from zero. You repeat yourself constantly — your job, your projects, your preferences — every single session, every single platform. Other tools either lock your memory to one provider or ask you to manage it manually.
Havril is a single memory layer that any model can read from and write to. Connect it once. Never repeat yourself again.
Havril is a free and open source project licensed under the MIT License. You are free to do whatever you want with it. Read the Contributing Guide before starting.
You could help continue its development by:
If you discover a security vulnerability within Havril, please send an e-mail to freedischthibaut@proton.me.
All reports will be promptly addressed, and you'll be credited accordingly.
