Overview
Monitor and analyze agent behavior with comprehensive observability tools. Use built-in LiveKit Cloud insights to view transcripts, traces, logs, and audio recordings, or collect custom data with data hooks for integration with external systems.
Agent observability is only available for LiveKit Cloud projects. It works for agents deployed to LiveKit Cloud and for self-hosted agents that connect to LiveKit Cloud media servers. It doesn't work with self-hosted media servers or entirely self-hosted deployments.
Where this data is stored depends on whether LiveKit Cloud collects it or your own code does. To learn more, see Data residency.
Observability components
Monitor agent sessions, collect metrics, and analyze behavior with these observability tools.
| Component | Description | Use cases |
|---|---|---|
| Agent insights | Built-in observability stack in LiveKit Cloud with transcripts, traces, logs, and audio recordings in a unified timeline for each agent session. | Viewing session transcripts, analyzing agent behavior, and debugging issues. |
| PII redaction | Remove personally identifiable information from session transcripts, audio, and telemetry before it's stored in LiveKit Cloud. | Reducing exposure of sensitive data in stored session recordings. |
| Data hooks | Collect session transcripts, reports, and metrics within the LiveKit Agents SDK for custom logging and integration with external systems. | Custom data collection, integration with external observability tools, and exporting data to your own systems. |
| OpenTelemetry traces | Export the OpenTelemetry traces from each session to any compatible backend, such as Langfuse , using the LiveKit Agents SDK. | Tracing agent behavior in third-party LLM observability tools. |
| Record audio and video | Capture session audio and video with LiveKit Egress and write the files to your own storage provider. | Recording video, or keeping media outside LiveKit Cloud. |
| Data residency | Where LiveKit Cloud stores the data it records, and where the data you collect yourself is stored. | Meeting regional storage requirements. |
Every component in this table provides information about a conversation: what the agent said, whether a tool call succeeded, or how the session performed.
Runtime logs provide information about the service running the conversation instead: whether a process crashed, what a build produced, or whether log lines reach your monitoring provider. This data belongs to the deployment rather than the session. To learn more, see the Agent deployment overview.
In this section
Agent insights
View transcripts, traces, logs, and audio recordings in LiveKit Cloud.
Build your own observability
Collect session data with the Agents SDK or LiveKit Egress and send it to systems you run.
Data hooks
Collect metrics, session reports, and recordings within the Agents SDK.
OpenTelemetry traces
Export session traces to any OpenTelemetry-compatible backend.
Record audio and video
Capture session audio and video to your own storage with LiveKit Egress.
Data residency
Where agent observability data is stored, and what you control.
PII redaction
Redact personally identifiable information from stored agent observability data.