### Answer When using Flex Mode in Conversation Flow, token counts can be significantly higher even with minimal content because all node instructions, transitions, and tool descriptions are compiled into a single prompt. [(1) ]() In Flex Mode, the entire conversation flow structure gets converted into one comprehensive prompt containing all Tasks and Tools, rather than sending only the active node's prompt to the LLM. This compilation happens regardless of whether you've added custom content to your nodes. The documentation specifically warns that "Flex mode can significantly increase your LLM costs. Because all node instructions, transitions, and tool descriptions are compiled into a single prompt, the total token count is much higher than in rigid mode." To reduce your token count, consider: - Using rigid mode instead of Flex Mode - Breaking your flow into smaller components so fewer nodes are compiled into a single prompt - Keeping node instructions concise to minimize token usage If you're seeing 1890 tokens on a "blank" agent, check if Flex Mode is enabled at either the Component or Agent level, as this would explain the high baseline token count.

​ To learn more, see the sources I considered: [[1] Flex Mode](https://docs.retellai.com/build/conversation-flow/flex-mode) ​

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

@AI Evy I am using Rigid mode

### Answer I don't have specific information in the available sources about baseline token counts in rigid mode for Conversation Flow agents or what causes high token usage with minimal configuration. The sources I have access to cover the Conversation Flow API structure [(1) ]() [(2) ]() [(3) ]() [(4) ]() and general prompt engineering guidance [(5) ]() , but don't explain the token accounting mechanics for rigid mode specifically. For troubleshooting high token counts in rigid mode, I'd recommend: - Checking your global_prompt field, as this applies to every node [(1) ]() [(4) ]() - Reviewing any tools or components configured at the flow level - Examining knowledge_base_ids if you have RAG enabled If you'd like more detailed guidance on this issue, you may want to: *Visit [Documentation]() * *Visit [Hire certified Retell partner]() * Would you like a general answer based on my knowledge instead?

​ To learn more, see the sources I considered: [[1] Get Conversation Flow](https://docs.retellai.com/api-references/get-conversation-flow) [[2] Update Conversation Flow](https://docs.retellai.com/api-references/update-conversation-flow) [[3] List Conversation Flows](https://docs.retellai.com/api-references/list-conversation-flows) [[4] Create Conversation Flow](https://docs.retellai.com/api-references/create-conversation-flow) [[5] Prompt Engineering Guide](https://docs.retellai.com/build/prompt-engineering-guide) ​

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