Memory should belong to the person it describes.
We believe AI systems deserve persistent, structured memory, and that this memory belongs to the person it describes. Esfis is a small, local engine that makes both true.
The name comes from the dried flower stem on the desk where the first prototype was written: two seed cups on one stalk, still holding after the season. Memory that outlasts the moment.
We started Esfis after building agents for three years and rebuilding memory for every one of them. The vendors' answer was always a private notebook inside their product. Ours is a store you own, that every product reads.
How we work
- Local first, then everything else
- If a feature needs our servers to work, we ask whether it should exist.
- Measured, not promised
- Every retrieval change runs against public benchmarks before it ships, and we publish the failures.
- Small on purpose
- One product, done properly. We would rather say no than spread ourselves across three.
- Write it down
- Decisions live in the changelog and the blog. If it wasn't written down, it didn't happen.
Careers
We're a small team based in Barcelona building the memory layer for every model and every harness. If that is the problem you want to spend the next few years on, we would like to hear from you.
Press
Logo, screenshots and a short description in the brand kit. For anything else write to press@esfis.com.