Hello! It sounds like you are building a great feature. The issue you are facing is that audio calls (WebRTC) must originate in the browser, not the backend. Your current axios.post request effectively tells Vapi to "start a call," but since the server has no microphone or speakers, the call has nowhere to go. Here is the correct architecture to securely gate your mock interview behind a subscription. The Solution: Backend Authorization, Frontend Execution Instead of "starting" the call on the backend, your backend should act as a Gatekeeper. It validates the subscription and then issues a temporary "pass" (a JSON Web Token) that the frontend uses to start the call securely.