LiveKit

LiveKit is a WebRTC-first platform for realtime video, voice, and data in your product. LiveKit Agents is its framework for conversational agents. Cartesia integrates in two ways: open source plugins and LiveKit Inference.

Quickstart

GitHub template for a basic LiveKit voice agent

Open source plugins

These are plugins that can be used with the open-source LiveKit Agents packages.

Python
Node.js

You can use Cartesia with Python LiveKit Agents by installing the livekit-plugins-cartesia package. If needed, you can the source code of the plugin in the /livekit-plugins/livekit-plugins-cartesia directory of the livekit/agents repo.

Ink 2 requires livekit-plugins-cartesia>=1.5.15.
Older versions will produce poor results without raising runtime errors.

Ink-2 natively handles turn detection, and the LiveKit plugin defaults to ink-2, which detects end-of-turn on its own; set turn_detection="stt" so the session uses those signals to know when the user stops speaking.

Setting up the Python plugin in your own agent

To get started, first install the main livekit-agents package as well as livekit-plugins-cartesia:

pip install livekit-agents livekit-plugins-cartesia

Then, use the Cartesia STT and TTS plugins when creating your agent:

from livekit.agents import AgentSession
from livekit.plugins import cartesia

session = AgentSession(
    stt=cartesia.STT(
        api_key=os.environ.get("CARTESIA_API_KEY"),
    ),
    tts=cartesia.TTS(
        api_key=os.environ.get("CARTESIA_API_KEY"),
    ),
    # ... llm, vad, turn_handling, etc.
)

Running a Python example

Check out /examples/other/cartesia.py in the livekit/agents repo for a fully working conversational agent.

# clone and setup livekit/agents
git clone git@github.com:livekit/agents.git
cd agents
uv sync

# run with required API keys:
# - CARTESIA_API_KEY
# - LIVEKIT_URL, LIVEKIT_API_KEY, LIVEKIT_API_SECRET
# -  some LLM provider's API key
uv run examples/other/cartesia.py console

You can use Cartesia with Node.js LiveKit Agents by installing the @livekit/agents-plugins-cartesia package. If needed, you can reference the source code of the plugin in the /plugins/cartesia directory of the livekit/agents-js repo.

LiveKit Agents is not feature complete in Node.js

Setting up the Node.js plugin in your own agent

To get started, first install the main @livekit/agents package as well as @livekit/agents-plugins-cartesia:

npm i @livekit/agents @livekit/agents-plugin-cartesia

Then, use the Cartesia STT and TTS plugins when creating your agent:

import { AgentSession } from '@livekit/agents';
import * as cartesia from "@livekit/agents-plugin-cartesia";

const session = new AgentSession({
  stt: new cartesia.STT({
    apiKey: process.env.CARTESIA_API_KEY,
  }),
  tts: new cartesia.TTS({
    apiKey: process.env.CARTESIA_API_KEY,
  }),
  // ... llm, vad, turnHandling, etc.
});

Running a Node.js example

Check out /examples/src/cartesia.ts in the livekit/agents-js repo for a fully working conversational agent.

# clone and setup livekit/agents-js
git clone git@github.com:livekit/agents-js.git
cd agents-js
pnpm i && pnpm run build

# run with required API keys:
# - CARTESIA_API_KEY
# - LIVEKIT_URL, LIVEKIT_API_KEY, LIVEKIT_API_SECRET
# -  some LLM provider's API key
node ./examples/src/cartesia.ts dev

# go to https://agents-playground.livekit.io/
# find your agent (make sure you're in the right org)
# then connect to it with WebRTC transport

LiveKit Inference

You can also use Cartesia’s models on LiveKit’s infrastructure. API keys and pricing are managed by LiveKit rather than Cartesia.

from livekit.agents import AgentSession, inference

session = AgentSession(
    stt=inference.STT(model="cartesia/ink-2"),
    tts=inference.TTS(model="cartesia/sonic-latest"),
    # ... llm, vad, turn_handling, etc.
)

import { AgentSession, inference } from '@livekit/agents';

const session = new AgentSession({
    stt: new inference.STT({ model: "cartesia/ink-2" }),
    tts: new inference.TTS({ model: "cartesia/sonic-latest" }),
    // ... llm, vad, turnHandling, etc.
});

Demo

LiveKit Cartesia Demo

Try out a simple voice assistant built with LiveKit Agents and Cartesia.

Simplified source code for the demo is available on GitHub in the cartesia-ai/livekit-agents-example repo.

Get Started

Text-to-Speech

Speech-to-Text

Tools

Integrations

Enterprise

Quickstart

Open source plugins

Setting up the Python plugin in your own agent

Running a Python example

Setting up the Node.js plugin in your own agent

Running a Node.js example

LiveKit Inference

Demo

LiveKit Cartesia Demo

Quickstart

​Open source plugins

​Setting up the Python plugin in your own agent

​Running a Python example

​Setting up the Node.js plugin in your own agent

​Running a Node.js example

​LiveKit Inference

​Demo

LiveKit Cartesia Demo

Open source plugins

Setting up the Python plugin in your own agent

Running a Python example

Setting up the Node.js plugin in your own agent

Running a Node.js example

LiveKit Inference

Demo