Deploy the LiveKit Egress service alongside your LiveKit server and ensure they share the same Redis instance for coordination.
Authenticate an EgressServiceClient with your API key and secret, then call startRoomCompositeEgress() with the room name, a layout (e.g., 'grid' or 'speaker'), and an EncodedFileOutput specifying filepath and an S3Upload block.
In the S3Upload block, supply your bucket name, region, and AWS credentials via environment variables (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY) rather than hardcoding them; set the output filepath to a template such as '{room_name}/{time}.mp4'.
Poll egressClient.listEgress({ roomName }) to confirm the egress job reaches the EGRESS_ACTIVE status before relying on the recording.
Stop the recording by calling stopEgress(egressId); the Egress service will finalize the file and upload it to S3, then transition to EGRESS_COMPLETE.
Subscribe to egress webhooks in your LiveKit server config (webhook.urls) to receive egress_started, egress_ended, and egress_failed events for automated post-processing.
Known gotchas
The Egress service requires its own host with a display (or Xvfb) for room-composite rendering; track-composite and track egress do not have this requirement.
If S3 credentials or bucket permissions are misconfigured, the job will fail silently at upload; validate IAM permissions (s3:PutObject on the target bucket) before starting production recordings.
Auto-egress configured in the LiveKit server config starts recording immediately when a room is created, which can produce many small files for short-lived rooms; use conditional auto-egress rules or start egress programmatically instead.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp