/docs/developerobservability

Logs, traces, and metrics

CloudWatch logs by default. OTEL traces when you opt in. Multi-agent dashboards on Team+.

CloudWatch logs

Every stdout/stderr line your agent emits lands in CloudWatch under /aws/bedrock-agentcore/{runtimeName}. The portal Logs tab tails this in real time. Retention defaults to 30 days; configurable on Department.

OTEL traces

Because we wrap your entrypoint with opentelemetry-instrument, every HTTP call, LLM invocation, and tool call is automatically traced. Traces export to AWS X-Ray by default. Bring-your-own OTEL endpoint (Honeycomb, Datadog, etc.) supported via custom OTEL_EXPORTER_OTLP_ENDPOINT env var.

Multi-agent dashboard (Team+)

/portal home shows all your runtimes side-by-side: 14-day sparkline of invocations, success rate, hours replaced, dollars saved. Click any runtime to drill into its individual dashboard.

Failure alerts

On the Settings tab per agent: enable email/Slack notifications for failed runs. Default is email-on-fail off, but most users flip it on within the first day.