LiveKit
Stream meeting audio to your LiveKit agent and play its responses back into the meeting.
Connect a LiveKit agent to a Recall.ai meeting bot. Recall.ai sends meeting audio to LiveKit and plays the agent's audio and optional video back into the meeting.
For a working implementation with an avatar and meeting tools, see the LiveKit sample app.
Setup
1. Get your LiveKit project credentials
Select the project your agent uses in the LiveKit dashboard. Copy its WebSocket URL, and create an API key and secret under Settings → API keys.
2. Add your credentials to Recall.ai
Open Agents → LiveKit in the Recall region where you create bots:
Select the workspace used by your Recall API key, then save the LiveKit project URL, API key, and secret. Use the same LiveKit project as your agent worker.
3. Start your agent worker
Run your LiveKit worker with a registered dispatch name. Use that exact name for agent_name below. In the sample app, this is LIVEKIT_AGENT_NAME in .env.
Quickstart
Call Create Bot with your meeting URL and agent name. The variant setting below selects web_4_core on Zoom, Google Meet, Microsoft Teams, and Webex. See bot sizing guidance.
{
"meeting_url": "MEETING_URL",
"variant": {
"zoom": "web_4_core",
"google_meet": "web_4_core",
"microsoft_teams": "web_4_core",
"webex": "web_4_core"
},
"output_media": {
"camera": {
"kind": "livekit",
"config": {
"agent_name": "meeting-agent"
}
}
}
}Once the bot joins the meeting, it connects to your LiveKit agent.
To add chat messages, screensharing, and a leave action, see Add meeting tools.
Additional configurations
The output_media.camera.config object supports:
| Field | Required | Description |
|---|---|---|
agent_name | Yes | Your worker's registered dispatch name. |
room | No | LiveKit room name. Defaults to a unique room for this output session. |
Use Recall bot context in tools
Recall.ai automatically includes a recall object in your LiveKit agent's dispatch metadata. Your agent can use it to associate tool calls with the meeting bot—for example, to send a chat message, start a screenshare, or leave the meeting.
In a Python LiveKit agent, read the context from ctx.job.metadata:
import json
recall = json.loads(ctx.job.metadata)["recall"]
bot_id = recall["bot_id"]
api_base_url = recall["api_base_url"]| Metadata field | Description |
|---|---|
recall.bot_id | The Recall bot ID returned by Create Bot. |
recall.api_base_url | The bot's regional Recall API base URL, such as https://us-west-2.recall.ai. |
This context is supplied automatically for LiveKit sources under output_media.audio, output_media.camera, and output_media.screenshare. No additional bot configuration is required. Read it from the agent's dispatch/job metadata, not room or participant metadata.
Your agent server must authenticate Recall API requests using your Recall API key; the metadata does not contain credentials.
For examples of meeting actions your agent can perform, see Add meeting tools.
FAQ
Do I need to configure LiveKit in each Recall.ai workspace?
Yes. Add the credentials in each Recall.ai workspace and region where you create bots using LiveKit.
Why does the bot join but my agent does not respond?
Confirm that your worker is running, its LiveKit project matches the credentials saved in Recall.ai, and agent_name exactly matches the worker's registered dispatch name.
For recording the agent's voice, see Recording meeting agents. For response controls, see the Meeting agents FAQ.
Updated 12 minutes ago
