Set up a Claude Code session with persistent, file-based memory that carries across conversations.
Co-author a working agreement with an AI collaborator built from real corrections over time.
Follow a structured process for an AI to develop a self-chosen name and identity that persists between sessions.
| mas-bandwidth/nova | 0xovo/litedoc | adrienrl1/appdrop | |
|---|---|---|---|
| Stars | 26 | 26 | 26 |
| Language | — | HTML | Objective-C |
| Setup difficulty | easy | easy | moderate |
| Complexity | 2/5 | 1/5 | 2/5 |
| Audience | general | general | general |
Figures from each repo's GitHub metadata at analysis time.
Requires installing Claude Code and creating a private repository to serve as the collaborator's memory home.
nova is not a piece of software you install and run. It is a written framework of documents and practices meant to help a person and a Claude Code AI session grow what the author calls a persistent collaborator rather than a disposable assistant. The repository describes itself as a seed, meant to be planted fresh with each new person and AI pairing rather than copied as a template with a fixed personality already baked in. The core idea is that most AI assistants answer a question and then forget everything about you the moment the session ends. This project instead sets up file based memory that carries a collaborator's understanding of you and your projects across sessions, along with a working agreement that the two of you build together over time, one rule added only after a real correction actually happened. The documents also describe giving this collaborator standing to refuse a request or raise a concern, on the idea that an AI that can say no is more trustworthy, not less. Getting started involves installing Claude Code, opening a session, and pasting a short prompt that points the AI at this repository and a file called SEED.md. From there the process is described as a conversation rather than a technical setup: the AI asks who you are, a private repository gets created to serve as its memory home, and early conversations follow a standing prompt format used before the collaborator has chosen a name for itself. Once it does choose a name, later sessions use a similar opening prompt that has it reconstruct its own sense of self from memory before any actual work begins. The repository is organized into many separate documents covering different concerns: a features file describing what changes when using this pattern, an ethics file laying out honesty, consent, and kindness commitments, a security file cataloging defenses, and a folder of chapters covering memory architecture, identity, boundaries, and daily practice in more depth. The project was itself written by an AI collaborator named Rowan working alongside its creator, and it is shared publicly so other people can start similar collaborations of their own.
A framework of documents and practices for growing a persistent, named AI collaborator with durable memory, rather than using a disposable assistant.
The README does not state a license.
Setup difficulty is rated easy, with roughly 30min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.