LiveKit Agents

LiveKit Agents is a Python and Node.js framework for building real-time voice AI agents that join calls as participants. It handles audio streaming, turn detection, and interruption over WebRTC, wiring together STT, LLM, and TTS providers into conversational agents. Integrations include OpenAI Realtime API, Deepgram, Cartesia, ElevenLabs, Google Gemini Live, and others. The framework supports tool use, multi-agent handoffs, vision input, and SIP telephony. It powers ChatGPT's Advanced Voice Mode and runs on LiveKit's open-source WebRTC server or LiveKit Cloud.