Skip to main content
Observability surfaces how Agent Graphs behave in each environment — sessions, tokens, telephony, execution logs, and billing allocation.

Usage metrics

Sessions, tokens, and telephony KPIs.

Session logs

Timeline, decisions, and captured variables.

Filtering

Slice by graph, environment, channel, and time.

Billing

Cost breakdowns tied to runtime usage.

​
What you can monitor

Metrics and logs reflect Agent Graph runs along the Design → Save → Build → Deploy path. They are unrelated to Agent Card A2A identity or external registry discovery.

​
Agent Graph runs

Every channel message, Chat API request, trigger webhook, or cron execution creates a session — one full run of a pinned build in an environment.
Filter logs by build version when debugging after a deploy — compare behavior between builds in the same environment.

​
Metrics at a glance

See Usage metrics for dashboard layout.

​
Debug workflow

  1. Open Observability from the workspace sidebar.
  2. Set time range and filter by Agent Graph, environment, or channel.
  3. Select a session to open the timeline.
  4. Inspect failed tool nodes and captured variables.
  5. Fix the graph or tool in Graph Studio, create a new Build, and redeploy.
PROD session logs may contain user PII from channel messages. Restrict observability access via RBAC.