livekit--agents/examples/voice_agents/grok

Grok Voice Agent API Example

This example demonstrates how to integrate xAI's Grok Voice Agents API with LiveKit Agents. It includes:

• Basic setup with xAI's voice-to-voice model
• Integration with built-in search tools so Grok can search x.com and the web

Quickstart

1. Set up environment variables

Note

All keys prefixed with LIVEKIT_ can be obtained from a project created on LiveKit Cloud or from a self-hosted LiveKit server instance.

export LIVEKIT_API_KEY=<your-livekit-api-key>
export LIVEKIT_API_SECRET=<your-livekit-api-secret>
export LIVEKIT_URL=<your-livekit-url>
export XAI_API_KEY=<your-xai-api-key>

2. Install dependencies

uv add "livekit-agents[xai,silero]" livekit-plugins-noise-cancellation

3. Run the agent

Option 1: Console mode

You can talk to Grok directly in your terminal:

uv run grok_voice_agent_api.py console

Option 2: Agents playground

LiveKit hosts a playground environment where you can connect to the agent running on your machine. First run the agent:

uv run grok_voice_agent_api.py dev

Then:

1. Open the playground environment in your browser: https://agents-playground.livekit.io/
2. Select the LiveKit Cloud project that's linked to your agent and click "Connect to "

Option 3: Run your own custom frontend

LiveKit offers a bunch of agent frontend starter templates across languages. Like Option 2, the first step is to run the agent on your machine:

uv run grok_voice_agent_api.py dev

Then:

Note

Whichever client platform you choose, to ensure your agent frontend and backend can connect to one another, use the same values for LIVEKIT_ environment variables with both your frontend and agent.

Clone and run agent-starter-react to interact with your agent in a browser

Or alternatively:

• Android
• Swift
• Flutter
• React Native
• ESP32
• Embed your agent directly into your existing website

4. Try it out

Ask Grok something that requires real time information:

"What is Elon Musk's most recent X post?"