LiveKit Agents

Overview

LiveKit Agents is an open-source framework for building realtime AI participants that run on servers and join LiveKit rooms. It is designed for voice, video, and multimodal agents that can listen, speak, see, call tools, exchange data with clients, and operate inside WebRTC or telephony sessions.

This is a strong fit when a Claude-adjacent team is building an actual realtime agent product instead of a chat-only workflow: voice support, phone support, client SDKs, media transport, job dispatch, testing, and deployment all matter at the same time.

Core Capabilities

Quick Start

Install the core package with the provider plugins you plan to use:

pip install "livekit-agents[openai,deepgram,cartesia]"

At minimum, a LiveKit-backed agent needs the LiveKit project/server connection:

export LIVEKIT_URL="wss://your-project.livekit.cloud"
export LIVEKIT_API_KEY="..."
export LIVEKIT_API_SECRET="..."

The repository documents local and production-oriented run modes:

python myagent.py console
python myagent.py dev
python myagent.py start
python myagent.py connect --room <room> --identity <id>

For AI coding agents, the upstream README recommends pairing:

• LiveKit Docs MCP server for current LiveKit docs and code examples.
• LiveKit Agent Skill for architecture guidance, handoffs, tasks, and testing patterns.

npx skills add livekit/agent-skills --skill livekit-agents

Example Shape

A minimal voice agent creates an AgentServer, starts an AgentSession, wires STT, LLM, and TTS providers, and starts an agent in a LiveKit room:

from livekit.agents import Agent, AgentServer, AgentSession, JobContext, cli
from livekit.agents import function_tool, inference


@function_tool
async def lookup_weather(location: str):
    return {"weather": "sunny", "temperature": 70}


server = AgentServer()


@server.rtc_session()
async def entrypoint(ctx: JobContext):
    session = AgentSession(
        vad=inference.VAD(),
        stt=inference.STT("deepgram/nova-3", language="multi"),
        llm=inference.LLM("openai/gpt-4.1-mini"),
        tts=inference.TTS("cartesia/sonic-3"),
    )

    agent = Agent(
        instructions="You are a helpful voice assistant.",
        tools=[lookup_weather],
    )

    await session.start(agent=agent, room=ctx.room)
    await session.generate_reply(instructions="Greet the user.")


if __name__ == "__main__":
    cli.run_app(server)

Use Cases

• Customer support voice agents that can answer, look up account state, and escalate to a human.
• Internal helpdesk or operations agents that join a room and guide an employee through a task.
• AI phone systems that receive inbound calls or place approved outbound calls.
• Multimodal assistants that combine audio, video, screen state, and client data APIs.
• Voice-first product copilots that need browser, mobile, or embedded client frontends.
• Production agent experiments that need tests, evals, staging rooms, and provider-switching before launch.

Source Review

Verified on 2026-06-18:

• The upstream README describes LiveKit Agents as a framework for realtime, programmable participants that run on servers and build conversational, multimodal voice agents.
• The README lists STT, LLM, TTS, realtime API integrations, job scheduling, WebRTC clients, telephony, data APIs, semantic turn detection, native MCP support, testing, and open-source deployment paths.
• The README documents the core install command and required LiveKit environment variables for the example agent.
• The repository AGENTS.md documents Python 3.10+, uv, unit/provider/eval test categories, and runtime commands for console, dev, start, and connect modes.
• The docs entry page describes LiveKit Agents as a realtime framework for voice, video, and physical AI agents, with start, build, server lifecycle, telephony, model, and deployment guides.
• GitHub reports Apache-2.0 licensing for livekit/agents, current active development, and a Python workspace with many provider plugins.

Duplicate Check

Checked current content/tools/, content/mcp/, content/skills/, content/agents/, guides, open and closed pull requests, and repository-wide content for LiveKit Agents, livekit/agents, livekit agents, voice AI agents, realtime voice agent, AgentSession, AgentServer, and livekit-agents. Existing entries cover adjacent voice, MCP, framework, and agent tooling, but no dedicated LiveKit Agents tools entry, LiveKit Agents source URL duplicate, or open duplicate PR was found.

Disclosure

Editorial listing. No paid placement or affiliate link is used. LiveKit Agents is Apache-2.0 open source; LiveKit also offers commercial cloud infrastructure and related services.