Pricing That Doesn't Meter You: Free, Pro, and Enterprise
Fernfly now has paid plans. Pro is $49 a month or $490 a year, with no per-message fee and no overage bill, because a model running on your own endpoint costs us nearly nothing per conversation. Here is the whole thing, including what changes on Free and the 80% lifetime discount going out to everyone who was here first.

Fernfly has been free since we opened it, on limits we set high on purpose and always meant to revisit. That day has arrived. We are introducing three plans: Free, Pro at $49 a month, and Enterprise.
The short version: Pro has no per-message fee and no overage bill. A quiet Sunday costs what a launch day costs.
Why flat, when everyone else meters
The argument Gaurav made in most AI work isn't reasoning has a pricing consequence we have never really spelled out.
If your assistant runs on a frontier model, every conversation is a metered API call, and your vendor's costs scale with your traffic. They have to pass that through. Per-message pricing isn't greed, it's arithmetic.
A fine-tuned Fern model is a different shape. It is small, it is yours, and it runs on an endpoint we already have warm. The tenth thousand request of the day costs us essentially what the first one did. Charging you per message would be inventing a meter to bill against a cost that isn't there.
So we don't. Pro is a flat subscription, and that is the whole pitch.
What you get
Free stays genuinely useful, not a demo timer. Point it at your site, train a model, embed the widget, and run it:
- 1,000 requests a day
- 15 training minutes a day
- 2,000 generated training examples a day
- Unlimited projects, and the embeddable chat widget
Pro, $49 a month or $490 a year, with the first week free:
- Unlimited requests, subject to a fair-use cap
- 500 AI fallbacks a day
- 60 training minutes a day, 20,000 generated examples a day
- Analytics, and retraining from the questions your assistant missed
- Email support
Enterprise is a conversation rather than a checkout button: on-premise deployment, dedicated training infrastructure, dedicated support, and limits sized to what you actually do. Talk to us.
What happens when you hit a limit
You get an email at 80% and again at 100%, so it is never a surprise. Limits reset at 00:00 UTC.
And if you are in the middle of something, click the "request grace credit" link in that email or simply hit the reply button. We will add grace credits to your account for the day. There is no form and no automatic top-up today because we want to talk to you about what you are doing and how we can help.
What changes if you are already on Free
You should hear this from us rather than from a lock icon.
The limits we launched with were placeholders. They were set high so people could experiment freely before there was anything to buy, and the code that seeded them has carried a note to lower them ever since. Here is the actual change:
| Before | Now | |
|---|---|---|
| Requests a day | 10,000 | 1,000 |
| AI fallbacks a day | 1,000 | 0 |
| Training minutes a day | 50 | 15 |
| Generated examples a day | 2,500 | 2,000 |
Two features also move to Pro: Analytics and Retraining. They are one loop rather than two features. Analytics shows you which questions your assistant missed; retraining is how you turn that list into a better model next week. Splitting them would have sold you half of it.
Free keeps what makes the product work: build a project, train a model, deploy it, embed it, and serve real traffic. Your first training run on any project is still free, on every plan.
If the new Free limits break something you had running, get in touch before you rearchitect around them. We would rather hear about it.
A thank-you to the people who were here first
If you already have a Fernfly account, watch your inbox over the next few days. We are sending every existing user a code for 80% off Pro, for life. Not the first three months. For as long as you keep the subscription.
You signed up when there was nothing to buy, no plans page, and no particular reason to trust us with a project. That is worth more to us than the revenue the discount gives up.
Why 80? Because it is large enough to be a real thank-you rather than a coupon, and because 80 is the number this company is built on: the 80% of AI work that isn't reasoning, which is exactly the part Fern was made for. It seemed like the right number to hand back.
AI fallbacks, and why Free has none
When your model meets a question it was not trained for, we can hand that question to a general-purpose model so your visitor still gets an answer. That call costs real money, per call, to us.
It is the one part of Fernfly that behaves like everybody else's pricing, which is exactly why it is metered separately and why Free has none. Without it, an out-of-scope question gets an honest "I don't know" rather than a silent bill.
Availability
All of this is live now. The plans are at fernfly.com/pricing, checkout is open, and the page reads its numbers from the same limits the product enforces, so what it promises and what the quota check does cannot drift apart.
Pro starts with a free week. Payment runs through Stripe, so card details never touch our servers, and cancelling is one button in the billing portal. If Enterprise is the shape you need, talk to us.
Anurag (aka `AnuRock`) leads tech and engineering at Fernfly, where he builds the platform that turns natural language into reliable tool calls. He aspires to be the dark lord of AI agents of the world one day.
LinkedIn