Tools not being triggered

Support Help
1

Hello folks,

For some reason, my agents are triggering the tools. They were working fine. Can’t understand what happened. They are not being triggered, and you may see that in the call transcription, if needed.

2

Based on the available information, there are two things you can concretely do inside Retell right now:

  1. Check / tighten your tool-calling prompt instructions
    For Single/Multi Prompt agents, Retell recommends giving explicit trigger conditions for each tool (function) in the system prompt, not just a vague description.(1)
    If your tools (e.g., qualification functions) used to be called and suddenly aren’t, it’s possible the LLM is no longer reliably inferring when to call them from a weak description. Strengthening the “Tool Usage Instructions” section (with clear “call TOOL_X when…” rules and exact tool names) is the recommended path when tools are not being invoked as expected.(1)

  2. Use the LLM Playground “Debug” flow to see what’s going wrong
    In Conversation Flow agents, when a node (including a tool/function node) doesn’t behave as expected, Retell’s guidance is to use the LLM Playground debug tools and then apply one of these fixes:(2)

    • Add fine-tuning conversation examples if the response itself is off.(2)
    • Add transition fine-tune examples if the node is not transitioning when it should (e.g., after a tool result).[^^2]
    • Split the node into multiple nodes so each node has a simpler, more deterministic behavior.(2)
    • Adjust LLM temperature and then use “Regenerate 10 Times” to test consistency.(2)

There is also a very similar report where custom “lead_calificado / lead_no_calificado” functions stopped being called; Retell support’s first step there was to look directly at the prompt section that should trigger the function and treat it as a prompting / configuration problem rather than a platform bug.(3)

Given that your tools “were working fine” and then stopped, the only documented, concrete next steps I can point you to are:

  • Tighten and make explicit the tool-calling instructions in your prompt for Single/Multi Prompt agents.(1)
  • Use the LLM Playground “Debug” button on the problematic responses/nodes, then iterate using fine-tune examples, node-splitting, and temperature adjustments until tool calls and transitions behave deterministically again.(2)

If you’d like help beyond what is documented, you can also reach out via these channels for hands-on troubleshooting:

If you want, I can also give a more general (non‑Retell‑specific) debugging checklist for LLM tool-calling behavior—just say so.