Everyday AI tasks for AI-native teams. At zero cost.
- Your users want to check an order, book a slot, file a request, update a record.
- Most teams build those flows on a large general model that reasons its way to a reply and charges for every word.
- Fernfly turns each request straight into the action instead.
Hoobert for WooCommerce, live on WordPress.orgThe same few things, all day long
They are the requests your customers and staff make constantly. Each one has an answer your system already knows, and an action your system can already take.
And the work that genuinely needs a large general model. Keep it there.
Write something new
Copy, replies, code. No single right answer, and the quality is the product.
Digest long documents
Reading a contract and telling you what matters needs the big model.
Weigh an open question
When the right answer depends on context nobody has written down.
Answer the request, or do it
“Where is my order 4821?” is one request. Here is what happens to it either way.
It reads your whole list of actions, works out what you meant, and writes an answer back.
A small model trained on your actions picks the right one, fills in the details, and your system runs it.
Function calling is the second one. It is not a smaller version of the first, it is a different job: pick the action, fill in the details, hand it back.
How it worksBuilt for this one job
We train, host and improve the small model that turns your customers’ requests into your app’s actions. That is the whole product.
Precise
Trained on your actions and nothing else, so it picks the right one and fills in the details. When a request is outside what it knows, it says so instead of guessing.
Instant
A small model doing one small job answers in under a second, which is the difference between a customer waiting and a customer not noticing.
Free to run
No fee per request. A busy Monday costs the same as a quiet Sunday, so the volume you wanted all along stops being a budget problem.
Apps are already shipping on it
We built Hoobert on Fernfly. A shop manager hits ⌘K anywhere in their store admin, types “refund order 62, damaged item”, and it runs the actual WooCommerce call. No menu-diving, no six screens, no paragraph of advice to act on yourself.
The model behind it is an ordinary Fernfly project: 28 store actions, one endpoint, trained from a template in an afternoon. Anyone can build the same thing on their own API.
Ask Hoobert: “refund order 62”Hoobert is a tool list and an endpoint. So is your app. Whatever your users currently click through, they could just ask for.
Build yours freeYour data stays yours
Runs where you want
Hosted by us, or inside your own network. The model is small enough for ordinary hardware, so it can sit where your data already is.
Optional third-party AI fallback
Your own model handles the request. The optional cloud fallback is one switch, per project, easily toggled off.
You own all of it
The conversations, the training data and the trained model belong to your account. Export or delete them at any time.
Just want a chat agent for your website?
Give us your address and we will read your site, train an agent that can use it, and hand you the snippet to paste. No API file, no account first.
Sized to you, not your message
Because a busy month should be good news.
Common questions
Start with one everyday task
Free to start, no credit card, and the first one is live in an afternoon.