Fern
the private assistantA branded B2B assistant that answers from your own knowledge. The frontend speaks only to your backend — and your backend is the control plane, the memory, and the AI gateway, all at once.
Your knowledge, answering back
Upload what your company already knows — documents, handbooks, contracts, wikis — and Fern turns it into an assistant that answers in your voice, citing your sources.
Every answer is grounded in an embedded vector store that lives on your own disk. Nothing is fine-tuned into a stranger's model; nothing is memorised somewhere you can't reach.
One gateway, no leaks
The browser never holds a key and never calls a model directly. Every request passes through your own backend — the single, auditable gateway between your data and any model.
Swap models, rotate keys, or run fully local: the frontend never knows the difference, and your keys are never exposed to it.
- 01
Grounded retrieval
A retrieval agent pairs a cloud or local model with a private vector store — every reply drawn from your own documents, not a model's guesswork.
- 02
Governed by an admin console
Users, assistant behaviour, capabilities, analytics, and the knowledge base — all administered from a considered admin surface, no config spelunking required.
- 03
Ingestion that keeps up
A dedicated ingestion worker indexes new material continuously, so the assistant's knowledge grows the moment your library does.
- 04
Deploys the way you work
A self-contained system that runs on your own infrastructure — piloted in the cloud, or deployed entirely on your servers for production.
Ferns predate flowers. They do not bloom or shout — they spread quietly by spore from the ground they are planted in, and endure where little else will.
Fern answers only from the knowledge already inside your organisation, unfurling one slow turn at a time from what you already own.