Lumi

About Lumi

Lumi is a persistent memory layer that spans ChatGPT, Claude and Gemini, so context you establish once is available in all three rather than trapped in whichever chat you happened to open. The problem it names is concrete and familiar: forty-five minutes spent giving an assistant project context, then a new conversation tomorrow that starts from zero; a switch from one model to another for a task it handles better, leaving the entire history behind; and usage limits burned faster on long projects because every message re-sends the whole conversation. Lumi's answer is a vault you write to from any of the three assistants and read from in any other, with memories organised into vaults and collections in whatever hierarchy you think in, and semantic search that retrieves by meaning rather than keyword. The efficiency claim follows from the retrieval model — instead of re-sending an entire history each turn it retrieves only what is relevant, which the site puts at up to 95% fewer input tokens on long projects, keeping costs flat as a project runs for months. On the privacy side it states that memories are end-to-end encrypted, never sold and not used to train AI, which is the necessary assurance for a product that by design holds the context you consider worth keeping. It is free to start.

Key Features

Tags