Fernfly adds a chat-style AI to your product, trained on your own features, running inside your app, with no per-message cost. Users ask for anything. Your app does the right thing.
Research and product
Fern and Fernfly share a bet: a small model trained for one job beats a large general one at that job. Fern is the research: compression to substantially fewer parameters, still in the lab. Fernfly is the product shipping today, and it runs whichever model fits, including off-the-shelf open weights.
The research is a bet on direction, not a dependency: Fernfly works now, without it. What the two share is the shape of the answer: a tiny, task-specialised AI that turns natural language into the right action inside your product, with no per-message bill and no data leaving your infrastructure unless you choose.
The lineage
Specialised, production-grade intelligence with substantially fewer parameters. Runs offline on phones, watches, and embedded chips.
substantially smaller models with measured quality retained, across language, vision, and embeddings.
A tiny AI you train on your own app, drop in with one URL, and run for free, for every user, forever.
Like ChatGPT for your app, but free and private
Describe what your app can do, train in minutes, drop in a chat box. Your users get a natural-language interface; you get a model that knows your product, runs inside your stack, and doesn't bill per message.
No per-message fee. Train once, run as many conversations as you want.
Trained on your own features and APIs, not generic internet knowledge.
Runs inside your app. User conversations never go to OpenAI or anyone else.
Real conversations improve it automatically. The more users talk, the better it gets.
See exactly how it performs on every version. Know when it works and when it doesn't.
A single API call. Add it to any app, web, mobile, voice, in minutes.
Free to start. No credit card. Set up in an afternoon.
