Voice Agents and MCPs Tutorial | devrel tutorial video made with ngram

About the video

This devrel tutorial video was made with ngram to walk developers through building a multi-channel persona system that keeps an ai agent's personality and memory consistent across voice calls, whatsapp, and other channels. the tutorial addresses identity fragmentation, the problem of a voice agent behaving differently depending on the channel, and proposes a gateway architecture to unify responses. it covers defining the agent's core identity through global prompts, integrating a rag system for centralized knowledge, and choosing a text-to-speech voice that matches the agent's written tone. the video explains deploying a central gateway server with a tool like openclaw to route traffic, connecting messaging platforms, and adding voice capability through bland ai with a 500-millisecond interruption threshold tuned for natural conversation. it also covers implementing cross-channel conversational memory so the agent recalls earlier text conversations during a follow-up call, removing the need for users to repeat themselves. the tutorial is aimed at developers building customer service bots, personal assistants, or voice-driven products who need a technical blueprint for a unified, multi-channel ai experience.