MokenLogic Memory
A per-person memory model that learns from finished sessions and returns sourced context.
About / MokenLogic
Subsidiary of Virtual Mind Holding (HKEX: 1520)Models improve every few months, yet what they know about the person using them resets with every new chat, app, or provider. MokenLogic builds the layer that keeps that context: learned over time, sourced, and usable by any model.
What we build / 01
A per-person memory model that learns from finished sessions and returns sourced context.
Planned access for other products to give their users the same continuity.
MokenLogic Harness for macOS, where memory, tools, and authorized actions work together.
How we work / 02
Memory is kept per person and is not tied to one model or provider.
Remembered facts keep their source, so a conclusion can be checked.
Incomplete memory is reported as incomplete, not filled in with a guess.
We share evaluations and research when others can check them.
Research / 03
Our research asks how an AI system can keep learning from one person’s work over months and years, and stay accurate as that person’s situation changes.
Learning from new sessions without losing what was learned before.
Keeping facts current, sourced, and specific to one person.
Letting remembered context guide authorized, reviewable actions.
Contact
For partnerships, research collaboration, or API access, reach the team through the contact page.