10 September 2026 · 4 min read
You spent an hour explaining your project to an AI. The next day, it no longer knows who you are. This gets framed as a technical limit. It's mostly a business model.
A captive memory, one per AI
Each vendor adds its in-house memory. Convenient, until the day you want to switch tools: it stays behind, opaque, unreadable elsewhere. The more you feed one AI's memory, the more attached to it you become. That's the intent, not an accident.
The real question isn't which, but whose
You're asked to choose which AI has the best memory. The right question is elsewhere: who owns yours. If it lives inside a model, it belongs to whoever runs the model. If it lives with you, it's yours, and any AI can come and read it.
A neutral layer above the AIs
Nawame doesn't talk and doesn't generate anything. It's the memory your AIs consult, through MCP, the open standard they already share. You plug in Claude today, Gemini tomorrow: the thread follows. And it keeps what actually matters, the reasoning and the decisions, not just raw facts.
Compartmentalized, and hosted in Europe
Everything is private by default. You share with whom you choose, to read or to write, without ever exposing your drafts. Your data is hosted in Europe and encrypted at rest. We won't promise the unreachable, we tell you what's true: you can export everything in Markdown or JSON, delete everything, and leave whenever you want.
Who it's for, honestly
If you're perfectly happy inside a single AI, you don't need us. Nawame is for people who move from one tool to another, who refuse to be locked in, and who want to stay in control of what they've built.