This section covers watching an agent after it's live — what conversations actually looked like, and what they cost.
Overview
Sessions and Traces gives you the detailed record of individual conversations: which tools ran, what knowledge was retrieved, and where an answer came from — the same trace view you'd use while testing, now for real production runs. Spending and Alerts and AI Metering cover the cost side: tracking what an agent has spent and how much it's being used, and setting limits or notifications so nothing runs away unnoticed.
What's in This Section
Topic | Covers |
|---|---|
Sessions and Traces | Inspecting real conversations and the trace behind each response. |
Spending and Alerts | Setting an agent's spend limits and notifications. |
AI Metering | Tracking token and request usage. |
Per-Agent vs. Workspace-Wide
Sessions and Traces and Spending and Alerts cover one agent's own Observability tab; AI Metering is the workspace-wide dashboard across all agents. For fleet-wide monitoring across every governed agent, and workspace-wide spending limits that apply to everyone, see the AI Governance and Libraries & Workspace Setup sections (in particular, AI Setup – Spending).