### Answer Retell AI counts tokens based on several components that are sent to the LLM during each response: What counts toward your token usage: - Your agent's prompt [(1) ]() - The transcript between the agent and user - Tool call history and results - Knowledge base content retrieved during the conversation [(2) ]() Token-based pricing: - Prompts up to 3,500 tokens follow the existing flat rate [(3) ]() - Prompts exceeding 3,500 tokens scale proportionally in cost (e.g., a 4,500 token prompt is billed at 4500 / 3500 × base rate) - The LLM prompt token limit is 32,768 tokens Why you might see surcharges: Even if your prompt is estimated at under 3,500 tokens, the actual token count includes the transcript and tool call history, which can push the total over the threshold during a call. [(1) ]() You can check your token usage in the call details within the dashboard.

​ To learn more, see the sources I considered: [[1] Token Surcharge when prompt tokens is under 3500](https://discord.com/channels/1186424376300679209/1442646663335510179) [[2] Will knowledge base use tokens?](https://discord.com/channels/1186424376300679209/1375421036291428452) [[3] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.