We build it for you, not a toolkit you assemble yourself
Starter is ready in 7 working days. Pro — phone, texts, and an agent that can act rather than only answer — is ready in 14. Underneath both sits the same reliability layer the platform has always run on: every customer reaches only their own data, a person signs off before anything destructive, and every answer is on the record.
A person reviews every build before it goes live. What's included at every tier is below.
Site chat
"Are you open Saturday?"
"Yes — 9 to 5, last booking's at 4."
Phone
"Are you open Saturday?"
"Yes — 9 to 5, last booking's at 4."
Text
"Are you open Saturday?"
"Yes — 9 to 5, last booking's at 4."
How it works
Every build — Starter or Pro — goes through the same four steps. Nothing is built until you've seen and approved the spec, and nothing is charged until you've seen and approved the agent.
1 · Describe it
Tell us what the agent should answer, what it should be able to do, and where your customers should meet it — your site, your phone line, your texts.
2 · Confirm the spec
We turn that into a written spec — the exact scope, the integrations, the tier and the price. You read it and approve it before anything is built.
3 · Watch it build
We build the agent against the confirmed spec and test it in a sandbox. You can follow along and try it yourself before a single customer does.
4 · Go live
It goes live on the day we quoted — 7 working days on Starter, 14 working days on Pro — and the balance is due when it does, not before.
Pricing
Every tier ships with the same reliability layer described below — scoped data, a human sign-off on anything destructive, a tamper-evident record. What changes is where your agent answers and what it's allowed to do.
Self-serve
An agent on your site that answers questions about your business.
+ $199/mo once it's live
Most common
An agent on your site, your phone line and your texts — and it can act.
+ $499/mo once it's live
By appointment
Inside your codebase, your deploy, or data that needs an agreement first.
from $999/mo once it's live
Every build splits into two payments: half when you confirm the spec, half when it goes live. Nothing is charged before you've approved what we're building.
What you actually get
This used to be the whole pitch. It still runs underneath every build — it's just not the reason to hire us any more. It's the reason it's safe to.
We bind your agent to whoever's signed in the same way for every customer — the model is never even shown the field that would let it reach someone else's order, account or record. There's nothing for a stray question, or a deliberately odd one, to talk its way into, because the door isn't in the room.
Anything that changes a record, charges a card or cancels something waits for a human before it happens. What that person is shown is not a summary of what the agent wants to do — it's the actual request, in plain terms. It only goes ahead once someone says yes.
Every conversation your agent has — what it was asked, what it answered, what it tried to do — is logged with its outcome. When an answer looks wrong, you flag it from the same view, not a support ticket that goes nowhere.
If we're not sure whether something already went through, we check before trying again — read the record back, never guess. What's below is a real one: five writes our side thought had landed once, that the provider's own records disagreed with.
each one is a 200 OK that did not mean what it said
Every privileged action your agent takes is written down in order and chained together, so the log can catch its own tampering. That's what "you get the transcripts" actually means here — not a promise, a receipt.
See one
This is AXONIC's own assistant, and it will answer questions about what we build, what it costs and how long it takes. Be clear about what it is: a taster. It holds no tools and can't do anything on anyone's behalf.
Your agent is the opposite. It reaches your real systems, under the rules you set, and you test it yourself in a sandbox — with its tools live and every guard armed — before you pay the second half of anything.
This bit needs JavaScript to answer. If you would rather just get on with it, start a build — you will be talking to your own agent in a sandbox before you pay anything beyond the deposit.
Start buildingBefore you start
Every build ships with the guardrails on this page: it only ever sees your own data, and anything that changes something waits for a person. When it doesn't know something, it says so and hands off instead of guessing. You get the transcript for every conversation, so you can see exactly what happened and flag it.
It's built for your business specifically — not a shared template — and it runs under your account from the day it goes live. The exact legal terms of that are in our terms of service .
Small changes — a new answer, an updated hour, a different tone — are covered by the monthly fee. Anything bigger, like a new integration or a new surface, goes through the same spec-and-confirm step as the original build.
You only pay the balance once you've tested the agent in a sandbox and approved it going live, so you're never on the hook for a build you haven't seen work. Once it's live, the monthly fee runs like any other subscription — cancel it and it stops.
The guardrails on this page — scoped data, a human sign-off on anything destructive — exist for exactly that reason. Beyond them, our terms of service set out what we're each responsible for. We'd rather you read that before you start than find out after something goes wrong.