LiveKit Agents

Overview

LiveKit Agents enables building voice AI agents that hold natural, real-time conversations. The framework handles the entire pipeline — capturing audio, transcribing, processing with an LLM, and synthesizing speech — with latency low enough for natural flow.

How It Works

1. Define agent — Set up STT, LLM, and TTS pipeline
2. Handle conversations — Process speech in real-time
3. Execute actions — Voice agent triggers tools
4. Deploy — Scale on LiveKit infrastructure

Use Cases

• Customer service — Natural-sounding voice bots
• Voice assistants — Custom domain-specific assistants
• Interview agents — Automated screening
• Language learning — Conversational practice partners

Getting Started

from livekit.agents.voice_assistant import VoiceAssistant
from livekit.plugins import openai, silero

assistant = VoiceAssistant(
    vad=silero.VAD.load(),
    stt=openai.STT(),
    llm=openai.LLM(),
    tts=openai.TTS(),
)

Example

User speaks: "What's the weather in San Francisco?"
Pipeline: STT → LLM (+ weather tool) → TTS → Audio response
Total latency: ~800ms

Alternatives

• Vapi — Voice AI platform (managed)
• Retell AI — Voice agent platform
• Bland AI — Phone call AI agents