Skip to content
Kitbay
Submit

voice-agent

MIT

Zoom's skill for Voice Agent V2 in Zoom AI Services, which runs a live spoken conversation over a WebSocket: it transcribes the user's microphone, runs an AI turn and streams back speech or text.

What the skill teaches

Zoom's skill for Voice Agent V2 in Zoom AI Services, which runs a live spoken conversation over a WebSocket: it transcribes the user's microphone, runs an AI turn and streams back speech or text. It documents the production route and session shape, PCM audio and voice activity detection, interruption handling and function tools, the order of events on the wire and how to recover from errors, and explains that the older V1 body needs a different subprotocol. It routes one-shot summaries, meeting media and transcription-only work to other skills.

Capabilities

  • Run a real-time voice conversation over WebSocket
  • Stream PCM microphone audio with voice detection
  • Handle interruptions and function tool calls