Agents do the work.
You approve what matters.
Give an agent a job your team repeats. It runs the whole thing, and stops for a person at the steps you choose.
Every run is recorded. Nothing important goes out unseen.
Solving everyone’s everyday task, connected in a series.
Two-time recipient of the UK Government compute grant · Microsoft partner
Supported by Google · NVIDIA · AWS · Aiven · NatWest
Three ways to work. One place to build them.
A Series you can call
Draw the steps once. It runs the same way every time, on a schedule or as an API call.
An agent you can talk to
Type to it, or pick up the phone. It hands you to the right specialist when the subject changes.
A person with the last word
Flag any step for approval. The run pauses, they get a task, and it continues when they say so.
Give it a job you are tired of.
It remembers how you do things.
Write it down once. Every agent carries it into every job, forever.
Not one agent. A team of them.
Real work is a chain of jobs. So you get a fleet: each agent does its part, hands off to the next, and stops for you where it counts.
Support desk
6 agents- Triage works out what it is about
- Lookup finds the order and the history
- Drafter writes the reply in your tone
- You approve anything with a refund
- Sender sends it and logs the ticket
- Reporter writes the weekly summary
Accounts payable
6 agents- Reader pulls the fields off the invoice
- Matcher matches it to the purchase order
- Checker totals, tax, duplicates
- You approve anything unusual
- Poster writes it into the ledger
- Chaser asks the supplier for what is missing
The phone line
5 agents- Reception answers and works out why they called
- Booking takes the appointment
- Billing answers the account question
- You take the ones it should not handle
- Notes logs the call and sends the summary
Inbound leads
6 agents- Enricher looks up the company
- Qualifier scores whether they are a fit
- Router assigns the right person
- Drafter writes the first reply
- Scheduler books the call
- Reporter keeps the pipeline honest
Hiring
6 agents- Screener reads every CV that lands
- Matcher scores it against the role
- Scheduler books the interview
- Notes summarises the conversation
- You make the yes or the no
- Writer sends the outcome, kindly
Bids and tenders
7 agents- Finder scans for new tenders daily
- Qualifier decides if it is worth bidding
- Extractor pulls out every requirement
- Drafter writes the compliant sections
- Compliance checks nothing was missed
- You sign it off before it goes
- Filer submits and archives the lot
Already running in hospitality, fixed-income trading, the public sector and outdoor safety. Fork one and change it, or start from nothing.
Type. Call. Or put them all on the same call.
Message an agent like a colleague. Ring one and talk to it. Or open a conference and have several agents and your team in the room at once, each speaking up when it is their turn.
Reception works out why you rang and brings in the specialist. Every word transcribed, every call logged.
Built for the teams who get audited.
Public sector, regulated finance, enterprise IT. Same questions, answered up front.
Start free. Scale without the token bill.
You pay in credits for work done, not per seat. Bring your own model key and the tokens cost you nothing. You only pay for the orchestration.
- ✓Unlimited with your own key
- ✓2 agents, 3 solutions
- ✓Public Stage rooms
- ✓Unlimited agents
- ✓1 private room, 1 site
- ✓10 solutions, no branding
- ✓Review gates and assigned tasks
- ✓3 seats included, then $39
- ✓Every solution and integration
- ✓Custom domains, top-tier models
- ✓White-label Stage
- ✓10 seats, 10 organisations
- ✓Run it for your clients
Unlimited seats and organisations. SSO, full audit log with export, private solution catalogue, on-premises option, a dedicated person who knows your name.
Annual saves 17%. One credit is about 2,000 words of thinking. A light job across five agents costs roughly 15 credits; a heavy one with tools and search, about 75.
Cheap models cost a quarter of a credit. Bring your own key and the model is free: you pay half a credit each time one agent hands to another.
The advanced work, built with you.
Not everything is a self-serve button. These we deliver as a custom implementation, our team alongside yours.
Agents that fan out
One agent plans, several specialists work in parallel, and their answers are checked and brought back together.
Agents that train themselves
Your corrections become lessons. Each agent trains a small model of its own, and stops needing an expensive one.
Developers: build one yourself at fernfly.com →A fleet built for your industry
We wire it, train it on your work, and hand it over. You own what comes out.
More on the edge. Models small enough to run on a phone, a laptop, or a device with no signal at all. Not here yet, and we will say so until it is.