Technologies for handling early media in VoIP communications with multiple endpoints are provided. A calling device sends an initial VoIP call request to multiple destination devices, or endpoints. The calling device then receives a provisional response from one or more of the destination devices that includes media streaming parameters regarding the destination device. The calling device creates a media context associated with the destination device that contains the media streaming parameters and stores the media context. The calling device uses the media context to establish a media connection with the destination. One of the destination devices returning a provisional response is selected to exchange early media over the media connection established with the destination device.