Memory that
belongs to you.

Most "personal AI" projects start from the model's capability and assume you will adapt to it. We think that is backwards. The intelligence is already remarkable, and improving fast — what's missing is everything around it that makes that intelligence land for one specific person.

Memory is the foundation of that. An assistant cannot be genuinely effective for you if it forgets you between sessions, or between tools. So we started with memory: a secure, portable layer that any AI tool can read and write, owned by the person it describes.

MyPenny is the first piece. The longer goal is a human-centered assistant built on top of it — one that becomes more effective the longer it knows you.

• The human is the point. Success is not "how capable is the AI" — it is "how much more effective is the person." Every decision routes back to that.
• Your memory is yours. Storage, export, and deletion are first-class. Encryption is per user. Data ownership is non-negotiable.
• Model-agnostic, always. Intelligence is whatever LLM you plug in. MyPenny never assumes one provider, and never locks you to one.
• Open where it counts. The memory engine is open-sourceable by design — the part that holds your data should be inspectable.