One mind. Every model.™
Persistent, self-hosted memory and retrieval for every AI you use.
Coming soon. Mnemnon is in active development — this repository is its future home. There's nothing to install yet; the architecture is settled and the build is underway.
Mnemnon™ (NEM-non) is a self-hosted, session-independent memory and knowledge-retrieval backend that serves all of your AI front-ends at once. Instead of each assistant starting from zero — and instead of your memory living on someone else's servers — Mnemnon gives Claude, ChatGPT, and any MCP- or REST-capable client one shared memory that persists across sessions, tools, and time, on infrastructure you own.
- Shared memory across every model. One memory layer behind Claude, ChatGPT, and any MCP/REST client — not another silo locked to a single assistant.
- Document RAG over your own corpus. Retrieval across your personal and work data that surfaces the right passage, with provenance.
- Graph-aware, long-term recall. Entities, relationships, and time-aware retrieval — so you get the current answer, not just a past one.
- Self-hosted and fully owned. Single-tenant, private, yours. Your memory never leaves your infrastructure.
- Headless by design. One clean boundary — REST and MCP — so any client can plug in.
Most "AI memory" is tied to a single assistant and lives in someone else's cloud. Mnemnon takes the opposite position: one memory, every conversation, always yours. Your AIs stop starting over, and your data stays home.
Early development. Architecture settled; build underway. Watch this space.
Mnemnon's source code is released under the Apache License 2.0 — you're free to use, modify, and distribute it under those terms.
That license covers the code only. Consistent with Section 6 of Apache-2.0, it grants no rights to the Mnemnon name, logo, or taglines — those are trademarks, addressed below.
Mnemnon™, and the taglines "One mind. Every model."™ and "Remember everything. Own all of it."™, are trademarks of Dax Davis (the "Marks").
Open-sourcing the code does not license the Marks. Without prior written permission, you may not use the Mnemnon name, logo, or taglines — or any confusingly similar variation — to name, brand, or promote your own product, service, fork, or distribution, or in any way that implies endorsement or affiliation.
Truthful, nominative reference is welcome: you may state that your project "works with Mnemnon" or "is built on Mnemnon," as long as you don't imply sponsorship or endorsement.
See TRADEMARKS.md for the full brand-usage policy.
© 2026 Dax Davis. All rights reserved, except as expressly licensed above for the source code.
Remember everything. Own all of it.™