How an AI answer is paid for
On the free plan your organisation gets 100 free AI answers a month. Once those are used, Vex keeps answering pay-as-you-go for as long as there are credits in your wallet, and hands over to a person when there are not. Paid plans have no free monthly allotment — every answer is billed from the first one — but at a lower rate than the free plan's, so credits go further.
An answer costs roughly $0.002 (about 2 credits). It is priced on the AI tokens the reply consumes, so the exact figure moves with the model and with how long the conversation and the reply are.
Credits are a prepaid balance in your organisation's Yaplet wallet. They are spent automatically as Vex writes replies, and they never expire. They are not added to or deducted from a Stripe invoice — the wallet is the source of truth. AI assist actions in the inbox (draft, summarise, rephrase) draw from the same balance.
What does not use credits
- The scripted steps of a workflow — messages, buttons, branches and collected fields. Three parts of a workflow do reach the AI and are billed like any other answer: an AI Bot step, which hands the conversation over to Vex; a message step with Let the AI say it switched on, which the AI rewrites for the visitor (it also counts toward that person's daily answers); and a workflow whose trigger is "User says", which needs an AI turn to recognise the intent.
- Storing or searching your knowledge. What costs is the AI work of the answer: writing the reply and, for shops with rules that find products, a small check that picks the rules matching a product question, billed with the answer. The Deep search depth (Brand settings → Advanced) costs slightly more than Standard, because it also searches with two more wordings of the question.
- Viewing reports, managing sources or editing settings.
- Changing a product's price or stock. Prices and availability are read live from your catalogue at the moment the answer is written, so changing them costs nothing — and the number in the AI's sentence can never disagree with the product card beside it.
Four things that now need credits
These used to run free for organisations with no subscription. They are now billed like any other AI generation, so without credits or a subscription they refuse with "AI generation needs credits. Top up your balance in Billing to keep generating."
- Translating your widget's text into other languages, from the widget's Languages tab.
- Vex's suggestion reports, which look for gaps in your knowledge. Reading reports you already generated is still free.
- Regenerating a visitor's memory by hand from the inbox. The automatic updates are unaffected — they ride on the conversation's own billing.
- Letting the AI write a help article or documentation page from a connected AI assistant. Writing it yourself costs nothing.
The free AI spend ceiling
If your organisation has no subscription and no credit balance, its AI usage is capped at $1 in any 24 hours and $5 in a calendar month, on top of the 100 free monthly answers. This is a safety net rather than a product limit: a real free account's busiest day costs a few cents.
At the ceiling, Vex hands new conversations to a person instead of showing your visitors an error, the dashboard's AI tools pause with a message naming the window, and the organisation's owners get one notification and one email that day. Adding credits or subscribing lifts it immediately; otherwise the daily ceiling returns within 24 hours and the monthly one on the 1st.
Three things that are cheaper than they used to be
- A knowledge base or documentation set is stored once, however many places read it. It used to be processed and stored separately for every widget that pointed at it — paid for twice, and the duplicate copies crowded out other sources when the AI ranked what to answer from.
- Moving a knowledge base to another brand costs nothing and re-processes nothing. Copying one costs nothing either, because every copied article arrives unpublished — you only start paying for the copy when you publish items at the destination.
- Refreshing a product feed only re-processes the products that actually changed. Keeping a large catalogue current is therefore far cheaper than it was, and an interrupted refresh repairs itself on the next run instead of leaving part of your catalogue unfindable.
Phone calls are billed separately
Calls are charged per minute rather than per answer, at $0.12 a minute. Phone calls also need the Voice AI module ($19 a month) — the per-minute charge is on top of it, and phone-number rental plus the one-time number setup fee are billed separately. The minutes draw from the same prepaid balance first, and anything beyond it is billed as metered usage on your subscription invoice. One caveat worth knowing: calls answered by a brand with no chat AI are billed and transcribed normally, but produce no AI answer-quality reporting.
Where to check usage
Open Brand → Vex → Reports. One screen holds the quality figures for a date range you choose, the table of AI-handled conversations, and the browsable list of every individual answer. For your credit balance and purchase history, go to Settings → Organization settings → Billing and look for the Yaplet Credits section.
AI setup has its own small allowance
An organisation with no subscription gets three runs in total of the "read my website and fill this in" action, and five different buttons draw on that same pool: the first-login wizard, the ＋ New brand dialog, Redesign with AI in the widget designer, Fill from website in Brand settings, and Smart import on the Knowledge page. A preview that saves nothing still uses one, because it still costs an AI call. Everybody, paid or free, is limited to five runs a minute.
The 250 indexed-item content allowance
Free accounts can hold 250 indexed items, shared across the whole organisation — every brand, knowledge base, documentation set, file, website page and question-and-answer entry counted together. One article is usually several items, so this is not "250 articles". The figure is shown on Brand → Knowledge as "212 of 250 indexed items used". At the cap, publishing new articles and adding new sources is paused until you remove something or upgrade. The same allowance applies if you bought individual modules instead of a plan — buying a module buys features, not content volume.
A free account can hold one of each of the things that are counted: one brand, one chat widget, one AI agent and one knowledge base. Each has its own allowance of one; they do not share a single slot between them.
A daily answer limit per person
One person — or one script pretending to be a visitor — cannot burn your credits by asking Vex endlessly. Each identity gets a fixed number of answers in a rolling 24 hours: 30 per website visitor, 10 per e-mail address and 30 per Messenger or Instagram user by default. Change the numbers per brand on Brand → Brand settings → Usage limits; an empty field means the default, and there is no off switch — enter a large number instead.
When someone reaches their limit, Vex sends one short closing message in their language (it never mentions limits), then pauses for that person for 24 hours. Your team sees a private "AI paused — daily AI limit reached" line in the conversation; the visitor does not. Their next message turns the conversation into a normal agent request, so nobody is left without a reply. Raising the number later does not shorten a pause that is already running. A normal support conversation never gets anywhere near these numbers.
Topping up credits
When your balance runs low, Yaplet emails you and shows an in-app alert so you can top up before it runs out. If it does reach zero and your free answers for the month are spent, Vex stops auto-replying and conversations fall back to your team — nothing is lost, visitors simply do not get an AI reply. The minimum top-up is $5 / 2,000 HUF and the maximum is $10,000 / 4,000,000 HUF. Topping up and managing the balance are owner-only.
Full details: Buy Yaplet credits for AI and voice usage.
Next step
See how the same content reaches every channel you pay for: Chat, inbox and phone all answer from the same brand knowledge.