How token pricing works
AI models read and write text in small pieces called tokens, and model providers charge by the token. OpenBot fingerprints your agents’ tokens to tell which model wrote them, and prices them at what that model costs you. The total is your model spend, which token pricing is based on.
See token pricing ratesHow a model call becomes a cost
Your agent
Inside every agent is an AI model. It writes in tokens, shown here as boxes, and they stream into OpenBot.
OpenBot
Model fingerprinting
OpenBot checks how the text splits into tokens, its word choice, formatting, and timing. Together these form a fingerprint, which OpenBot matches to a known model.
Rate
OpenBot uses the model’s public price. If you pay a different rate or host the model yourself, set your own pricing formula instead.
Cost of this request
OpenBot totals costs by model, Bot, and user. Token pricing is based on your total, and the same totals power spend controls.
Questions about token pricing
Something else? Talk to an engineer
Do I still pay my model providers?
Yes. You pay your model providers directly. OpenBot doesn’t resell model access or mark up tokens.
Is the estimate exact?
It’s an estimate, based on the tokens OpenBot measures and each model’s price. Your provider’s invoice is still the final word on what you owe them. If you pay something other than the public price, set a pricing formula so the estimate uses your rates.
Can I get exact pricing instead of an estimate?
Yes. If your company sends model requests through an AI gateway, OpenBot can connect to it and price each request from the exact usage the gateway records. Talk to us to set it up.
What about models without a public price?
Models you host yourself or buy under a private agreement may not have a public price. Set a pricing formula for them and OpenBot prices their tokens with it.