Skip to main content
Agent Builder v2 is in open beta. Features and limitations described here are subject to change.
Each Agent Builder v2 agent has an Observability tab with request-level metrics, a request log, and usage data, plus server and deployment logs for each environment. Use these views to confirm a deployment is healthy, debug a failing request, and track who’s using the agent and what it’s costing.
This page covers monitoring for an individual Agent Builder v2 agent. For organization-wide observability—captured LLM calls across your account, usage and spend, and export APIs—see Observability overview.
To open an agent’s observability views, log in to Control Center, go to the Agents tab, select your agent, then select Observability. The Observability tab has three views: Performance, Events, and Usage. Server and deployment logs are available separately, from each environment’s Logs tab on the agent’s Deploy page.

​
Review request performance

The Performance view shows request-level metrics for the agent’s container over a selected time range, including request volume, latency, and error rate. The Performance view showing requests, error rate, average/p95/p99 response time, and a response-time chart over a selected date range Use the time range controls to select a preset window or a custom start and end date. The metrics update to reflect the selected range; a custom range always covers the full UTC days it spans, not just the hours within them. Watch this view after a deployment or a promotion to confirm the agent is handling traffic normally. A rising error rate or a latency spike that starts right after a deploy usually points to the new version, not to an unrelated dependency.

​
Inspect individual events

The Events view lists log events from the agent’s container, each with an ID, timestamp, level, type, service, and message. Select a row to see more detail about that event.
Agent Builder v2 agent events don’t appear in the AI Studio or Control Center event logs. They’re only visible in the agent’s Observability tab.
The Events view showing successful and error events per day, with the event log table below An event is a single log entry from the agent’s container, not a trace of everything your application code did while handling a request—it doesn’t break out individual model calls, tool calls, or retries. For model-level usage, see Track usage by team and user below. Use the Events view to confirm an agent is logging activity, to find out when a specific event occurred, and to check its level—for example, whether it logged an error.

​
Track usage by team and user

The Usage view shows views, unique users, model requests, and spend for the agent’s production deployment over a selected time range, with spend broken out per model and a table of the top authenticated users.
The Usage view only reports on the agent’s production environment. It doesn’t include traffic from staging or other non-production environments.
The Usage view showing total views, unique users, model requests, spend, a spend-per-model chart, and a top users table Filter by team or seat type to narrow the numbers to a specific group. Use this view to see who’s using the agent, how much it’s costing, and which models are driving that spend.

​
Tail server and deployment logs

Each environment has its own Logs tab, available from the agent’s Deploy page: select an environment, then select Logs. The logs toggle between two views:
  • Server—streams output from the running environment, so you can see what your application is printing in real time.
  • Deployment—shows per-step build and deploy log lines with timestamps for a snapshot, so you can see exactly where a deployment succeeded or failed.

​
Next steps

Continue setting up and learning about Agent Builder v2 with these guides.