Bots work perfectly when I test them 1 by 1. But when I put them together in a squad, something breaks. Example: Bot is skipping checking Go High Level calendar and just spits out a bunch of fake times.

Suggestion: Check if your squad configuration explicitly defines the tool usage for each assistant. In your squad payload, ensure the assistant responsible for checking the Go High Level calendar includes the correct tool ID and a clear system prompt. For example:

json
{
  "members": [
    {
      "assistant": {
        "name": "Calendar Checker",
        "model": { "provider": "openai", "model": "gpt-4o" },
        "toolIds": ["GO_HIGH_LEVEL_CALENDAR_TOOL_ID"],
        "firstMessage": "Let me check the calendar for available times.",
        "firstMessageMode": "assistant-speaks-first"
      }
    }
  ]
}

Test this configuration in the Vapi dashboard and review the call logs to confirm the tool is being called as expected in the squad context. Sources: - [Property management routing squad](https://docs.vapi.ai/squads/examples/property-management) - [Debugging voice agents](https://docs.vapi.ai/debugging)

Hi Yssa Mae Amper, If you experience issues with a custom tool not triggering in your Vapi assistant, consider these steps: 1. **Check Assistant Prompts**: Ensure the exact tool name is used in prompts. For a tool named

get_weather

, use

get_weather

directly instead of variations like

weather_tool

. 2. **Verify Tool Schema**: Ensure that all required parameters are defined in your tool's schema. Example:

json
   {
     "name": "get_weather",
     "parameters": {
       "type": "object",
       "properties": {
         "city": {
           "type": "string",
           "description": "City name for weather lookup"
         }
       },
       "required": ["city"]
     }
   }

3. **Enable Schema Validation**: Add

"strict": true

in your tool configuration to catch errors early:

json
   {
     "name": "get_weather",
     "description": "Get current weather for a city",
     "strict": true,
     "maxTokens": 500
   }

4. **Response Format Errors**: Ensure your webhook returns the correct JSON structure and HTTP 200 status for all responses. Example success response:

json
   {
     "results": [
       {
         "toolCallId": "call_123",
         "result": "Your response as single-line string"
       }
     ]
   }

Inspect call logs for "Schema validation errors" or "Token truncation warnings" to diagnose further issues. You can [find more details here](https://docs.vapi.ai/tools/custom-tools-troubleshooting).