By the end of the guide, you know which plan fits your volume, what may come on top of the subscription and how to monitor usage once the chatbot is live.
Before you start
- An estimate of how many questions your customers ask each month: calls, emails or messages on repetitive topics.
- The list of documents and pages the chatbot must know, with their approximate size.
- The channels you want to cover: website, WhatsApp Business, Telegram.
- The number of chatbots you need: one per website, per WhatsApp channel or per clearly separate use.
1. Count replies, not visitors
Each chatbot reply uses tokens: the question, your instructions and passages from your documents are read again to write it. Chatyx estimates assume about 15,000 tokens per AI reply, a figure measured on an agency chatbot. A four-turn conversation therefore uses roughly four times that amount.
On the paid plans, that gives per month — Starter: 3,000,000 tokens, about 200 AI replies; Growth: 9,000,000 tokens, about 600 AI replies; Scale: 30,000,000 tokens, about 2,000 AI replies; Enterprise: 80,000,000 tokens, about 5,300 AI replies. These are orders of magnitude: a long reply or a drawn-out conversation uses more, a simple question less.
To estimate your needs, start from the repetitive requests your team already handles, the ones an AI chatbot for small businesses takes on. Fictional example: a workshop receiving twenty questions a day, with three exchanges on average, reaches about 1,800 replies a month. Choose the plan whose volume covers that figure with some margin.
2. Check your source volume
The chatbot answers from the sources you give it: web pages, sitemaps, PDF, DOCX, TXT, XLSX or CSV files. Their number is limited by plan: 1 on Free, 5 on Starter, 20 on Growth, 100 on Scale, 5,000 on Enterprise.
Indexing a document uses storage tokens, and files take up space: 10,000 tokens and 100 MB of files on Free, 500,000 tokens and 1 GB of files on Starter, 2,500,000 tokens and 5 GB of files on Growth, 10,000,000 tokens and 20 GB of files on Scale, 30,000,000 tokens and 100 GB of files on Enterprise. A website of a few dozen pages fits in the smaller plans; a catalogue of several thousand items means checking both limits.
A resynced page only uses tokens again for the content that changed. An unchanged document is not counted again at every update.
3. Count chatbots and channels
Each chatbot takes a slot in your plan: 1 on Free, 2 on Starter, 5 on Growth, 10 on Scale, 100 on Enterprise. Connecting WhatsApp Business creates a chatbot dedicated to that channel: plan its slot on top of your website chatbot.
On WhatsApp, Meta bills delivered messages under its own rates, on top of the Chatyx subscription. These fees do not go through Chatyx: check Meta’s pricing for your country.
There are no setup or development fees. A Shopify store or a real estate agency page is connected from the app, with no service to order.
4. Choose the plan and billing
Current prices are as follows — Starter $21 a month, or $14.92 a month billed annually; Growth $43 a month, or $30.75 a month billed annually; Scale $87 a month, or $62.42 a month billed annually; Enterprise $162 a month, or $116.58 a month billed annually. Any taxes appear in the payment summary.
Beyond volume, plans differ in support (email from Starter, priority on Growth and Scale, VIP 24/7 on Enterprise) and in fast replies, available from Growth: the chatbot then starts answering sooner.
The Pricing page remains the reference: it is updated with the catalogue, and its calculator uses the same usage assumption.
5. Test on the free plan, then adjust
The Free plan gives one chatbot, one source and a test allowance of 250,000 tokens, about fifteen AI replies. It is there to check answer quality on your own documents, not to absorb real traffic.
Once the chatbot is live, follow its token usage in the dashboard for two or three weeks. Compare it with your estimate, then adjust the plan: the usage measured on your own conversations, not an average, should drive the choice.
Check the result
- The estimated monthly replies fit the chosen plan, with some margin.
- The planned sources respect the plan’s number and storage limits.
- Each channel has its chatbot slot, and Meta’s WhatsApp fees are budgeted separately.
- Actual usage is checked after a few weeks and compared with the estimate.
Troubleshooting
I do not know how many questions I will receive.
Start from the requests your team already handles: calls, emails, messages on repetitive topics. Failing that, start with a smaller plan and check usage in the dashboard for a few weeks before changing.
The allowance drops faster than expected.
Long conversations, detailed replies or many passages read again raise the usage per reply. Check the longest conversations in the dashboard, tighten the instructions if the chatbot answers at too much length, then reassess the plan.
WhatsApp fees do not appear on my Chatyx invoice.
That is expected: Meta bills WhatsApp messages under its own rates, outside the Chatyx subscription. Check your Meta account for these fees.
Need help with your setup? Email support@chatyx.fr.
Further documentation
Run your own test
Start with a clear scope and a few reliable sources. Check the answers before expanding to visitors.
Open Chatyx