Chatbot API token costs: the metered part
Model providers bill by the token, and they quote input and output tokens at separate rates per million. Each question, each retrieved passage and each answer uses tokens, so spend follows traffic. Busy assistants, or ones hit by a scraper, cost more, and without a limit the first sign is the invoice.
The engineering answer is a hard token budget: the assistant gets a spending cap, and unusual usage triggers an alert before it adds up. Anthropic's pricing page(opens in a new tab) notes that its newer tokenizer produces about 30% more tokens for the same text, so a model upgrade can change spend even when traffic does not.

| Cost driver | Why it moves the bill |
|---|---|
| Traffic | More conversations means more tokens |
| Prompt length | Instructions and retrieved passages are sent with every question |
| Answer length | Output tokens are priced separately from input |
| Model choice | Each model has its own rates, and tokenizers differ |
Hidden costs AI assistant teams miss: re-indexing
The index is a snapshot, and your business changes. Prices move, policies update and products launch, and each change widens the gap between what is true and what the bot retrieves. Re-indexing is a recurring job, and it belongs on a trigger tied to content changes, not on someone's memory.
Budget for it as an event. Publishing or editing a source page queues the re-index. Repeating last quarter's pricing with full confidence is usually a re-indexing failure, not a model failure.
Monitoring: the cost of a wrong answer
One confidently wrong answer to a paying customer costs trust and sometimes money. Monitoring means scheduled test runs against known answers, a sample of live transcripts and a review of every model change. Ours starts at $497 a month, on the AI Operations tier, as part of Production-Safe AI Management.
A chatbot cheap enough to skip monitoring is a chatbot whose costs you have not counted yet. What the checks involve is covered in AI chatbot maintenance and monitoring.

Our published prices for a knowledge-base assistant
We publish two prices, a $4,500 fixed build for a knowledge-base assistant and monitoring from $497 a month, which comes with every build. Model token charges are billed on top, passed through at the provider's rate and capped by a token budget. Ask any vendor the same question, because tokens follow your traffic.
For how the assistant answers from your documents, see RAG chat with your documents.
| Item | Price |
|---|---|
| Knowledge Base Assistant, base build | $4,500 fixed, one time |
| Production-Safe AI Management | From $497 a month, required with every build |
| Token spend | Billed on top at the provider's rate, with anomaly flags and a token budget cap |
Questions about AI chatbot maintenance cost
How much does an AI chatbot cost to run each month?
It depends on traffic, prompt length and the model, so a fixed figure without your usage would be a guess. Estimate it as conversations a month, times tokens per conversation, times the provider's rate per million tokens. Then add re-indexing and monitoring.
Can I cap what a chatbot spends?
Yes. Set a hard token budget, and spend stops at the limit while an alert flags unusual use before the invoice does. Our builds include token budget caps and spend reports.
Why does a model upgrade change my bill?
Each model has its own rates, and tokenizers differ. Anthropic notes that its newer tokenizer produces about 30% more tokens for the same text. Test cost on the new model with your own questions before you switch.
