Most AI assistants require you to send your data to the cloud. Nova is built on a simple, powerful idea: your AI should work for you, on your hardware, under your control.
Nova is an open-source AI secretary that runs entirely on-premise. It brings the power of a dedicated assistant to your workflow while guaranteeing that your data remains private. Period.
Nova acts as a true secretary, with privacy as its most important duty.
- 🔒 Private & On-Premise: This is Nova's core promise. Run it on your own servers. Your data, conversations, and models never leave your infrastructure.
- 👀 Transparent & Autonomous: Nova manages its own task list on a kanban board. You can watch it work, provide guidance, and have the final say, offering the perfect blend of autonomy and control.
- 🧠 Remembers & Improves: With a persistent graph memory, Nova learns from your interactions. If it lacks a skill, it can request a new tool from its developers, creating a foundation for continuous improvement.
- 🔌 Adaptable Skills: While it comes with core secretarial skills (task management, email processing), Nova can be extended with new abilities via MCPs to fit your specific needs.
This is a project driven by a vision, but it is still in its early stages. Many of the features described here are under active development. We are looking for enthusiastic developers and contributors to help us turn this proof-of-concept into a robust, real-world application.
If you are excited by the idea of a private, collaborative AI, we invite you to join us. Your contributions can help shape the future of Nova.
Get your own AI secretary running in under a minute.
- Prerequisites: Ensure you have
dockeranddocker-composeinstalled. - Run Nova:
docker-compose up -d
- Open Your Browser: Navigate to
http://localhost:3000and give Nova your first task.
- Want to see how it works? ➡️ Read our Architecture Deep Dive.
- Want to contribute? ➡️ Check out the Developer Setup Guide.
- Want to see the full vision? ➡️ Explore our Documentation.
Backend: FastAPI, LangChain/LangGraph, SQLAlchemy, PostgreSQL, Redis Frontend: Next.js, React, TailwindCSS AI: Google (default), configurable for local models via LiteLLM