This is a really common setup issue with Vapi AI, so you’re not alone here. The “Messaging” URL only listens to events, it won’t let you control the AI before it speaks. If you want AI → Backend → AI flow, your webhook needs to be in the main Assistant/Voice Server URL, the one tied to the assistant-request flow, not under Messaging. That’s the only place where Vapi pauses and waits for your backend to send the firstMessage and system prompt. The right flow should be: call starts → Vapi hits your server → you inject your logic → you return the assistant → AI speaks. If you want, you can send me a private message and I’ll show you exactly where to place it and how to structure the response so it works smoothly. @Puffy