AI Agent

What the assistant costs

Every reply records its model and tokens; the analytics add them up.

A smiling barista behind the counter
AI Agent

Each assistant reply stores which model answered and how many tokens it used, priced at the model’s published rate. Analytics → AI support shows the total for the period; the inbox shows the cost of each reply under the message. Nothing is estimated.

The model is chosen in AI Agent → Configure. A harder question can escalate to a stronger model for that one reply; the breakdown is kept per reply. On the Free plan the AI Agent answers on the standard model and does not escalate: it answers, or hands the conversation to a person.