Getting Started

LiveKit Agents allow you to quickly create an infinitely customizable voice AI agent in Python.

Quick Start with CLI

Using the CLI you can start talking to your agent in 4 steps:

  1. Install the LiveKit CLI: brew install livekit-cli OR curl -sSL https://get.livekit.io/cli
  2. Authenticate with LiveKit Cloud: lk cloud auth
  3. Clone the agent boilerplate: lk app create
  4. Install dependencies and run your agent: cd your_agent_directory && uv sync && uv run src/agent.py download-files && uv run src/agent.py console

Congrats, you’re now speaking with your agent!

  • All of your agent code lives in src/agent.py
  • See all of the different agents you can build in the examples repo

đź“– Docs quickstart link (includes more copy / paste start from scratch stuff)


MCP Servers

LiveKit also offers a few different MCP Servers that help coding assistants write code. If you add them to Cursor or Claude Code, the assistants will automatically reference documentation and examples when they’re building agents to keep hallicinations to a minimum.

Python Examples MCP Server

https://livekit.mcpbin.org/mcp

Install with Cursor

Developer Documentation MCP Server

https://docs.livekit.io/mcp

Install with Cursor


Frontend Integration

When you’re ready to talk to your agent from the web or mobile app rather than through the console:

  1. Clone the frontend boilerplate: lk app create
  2. Select the template you’d like to clone, ex: agent-starter-react
  3. Move into your directory, then: pnpm install && pnpm dev
  4. Visit the link to talk to your agent
  5. You can also just use a prebuilt sandbox as your frontend

  • Python Agents: We recommend that you build agents with this framework.
  • Python Examples: See what you can build with these examples.
  • Node.js Agents: This lightweight Node.js framework doesn’t yet include all the features that come bundled with the Python Agents framework.

Billing Information


Deployment

When you’re ready to deploy your agent to the cloud:

  • Deploy to cloud w/ livekit agent create from within your agent directory.
  • Test your deployed agents with LiveKit Sandbox or the frontend you cloned above.