### Answer Based on your debug output, your tool call is actually working correctly—the tool result shows a successful response with

{"status":"dynamic variable extracted","variables":{...}}

. [(1) ]() The "tool accepted" message you may have seen previously is not a standard part of the tool call result format. What matters is that: 1. The

tool_call_id

is present and matches 2. The tool returned a valid result with your extracted variables If you're using a Custom LLM WebSocket, you can send tool call invocations and results to Retell to have them weaved into the transcript. The Tool Call Result Event requires: - `response_type`: set to

"tool_call_result"

- `tool_call_id`: the globally unique tool call id - `content`: the result content as a string Your tool appears to be functioning properly since it's returning the extracted dynamic variables. If you're experiencing different behavior than before, check: - Whether any configuration changed on your server that handles the tool calls - If you have server-side validation that might be rejecting requests (a 422 error can occur if your server has validation that rejects the request) [(2) ]() Could you clarify what specific behavior changed? Is the agent not acknowledging the tool result, or is there an error occurring?

​ To learn more, see the sources I considered: [[1] LLM WebSocket](https://docs.retellai.com/api-references/llm-websocket) [[2] Custom Tool call fails](https://discord.com/channels/1186424376300679209/1400798045930651718) ​

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

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

### Answer I don't have specific information about Make.com integrations with Retell AI in the available sources. However, based on your debug output from the previous message, your tool call is completing successfully on the Retell side—the tool result shows the variables were extracted properly. If your Make.com automations aren't triggering, the issue is likely on the Make.com/webhook side rather than Retell. Here are some things to check: 1. Webhook configuration – Verify the webhook URL in your tool is correct and Make.com is listening 2. Webhook status – Ensure your Make.com scenario is active and not paused 3. Request format – Confirm the payload format matches what Make.com expects 4. Make.com logs – Check Make.com's execution history to see if requests are arriving but failing If you're using Retell webhooks (separate from tool calls) , you can subscribe to specific events like

call_analyzed

to trigger external automations. [(1) ]() Could you share more details about how your Make.com automation is connected—is it via a tool call HTTP request, or via Retell's webhook events?

​ To learn more, see the sources I considered: [[1] Register Phone Call](https://docs.retellai.com/api-references/register-phone-call) ​

### Answer I understand the frustration. There have been reports from other users experiencing similar issues with tool calls suddenly not working as expected. [(1) ]() A few things that may have contributed to changes: Recent LLM model changes: - GPT-4o was discontinued recently, and agents were migrated to GPT-4.1 - One user reported: "Previously 4o, but you (or OpenAI) discontinued it a couple days ago" What to check: 1. Verify your LLM model – If your agent was using GPT-4o, it may have been automatically switched. GPT-4.1 is recommended as a replacement. 2. Review your function prompts – The prompting for when to trigger functions may need adjustment for the new model. As one support member noted, "the problem must be prompting if you can send me prompt where you call the function." 3. Check for platform incidents – There was a recent incident affecting tool call parameters: [(2) ]() 4. Temperature settings – For reliable function calling, use lower temperature (0.0-0.3) for more deterministic behavior. [(3) ]() Since your tool result shows the data is being extracted correctly on the Retell side, but Make.com isn't receiving it, the issue may be in how the webhook is being triggered after the tool completes—not the tool call itself. Could you confirm which LLM model your agent is currently using?

​ To learn more, see the sources I considered: [[1] Functions stopped being called out of nowhere](https://discord.com/channels/1186424376300679209/1436407466404479108) [[2] URGENT: Tool calls are not passing in required parameters](https://discord.com/channels/1186424376300679209/1403199812077092945) [[3] Configure LLM Options](https://docs.retellai.com/build/llm-options) ​

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