Deepgram co-developed the Real-Time Context Standard open protocol with the Pipecat team to pass audio context metadata—such as whispers, tone, and model guidance—between voice agent components.