Usage Monitoring & Cost Visibility for Admins

**Value proposition**
As Agents move into production, admins need visibility into how agents are being used — which conversations are consuming the most tokens, where context windows are being saturated, and where costs are spiking — so they can optimize and govern usage.

**Expected outcome**
A consumption API & dashboard in the agent management view shows per-conversation token usage, round counts, and warnings for high-cost conversations. Superusers can sort and filter by usage metrics. Operational tracing provides error visibility across customers for diagnosing issues.

**Key user stories**
- As an admin, I want to see which conversations are consuming the most tokens so I can identify runaway agent loops or inefficient workflows.
- As an operator, I want to be alerted when a conversation's context window is near saturation so I can intervene before costs escalate.

**Expected Phases**
- API for monitoring token usage per agent and conversation
- Index with all data available for analysis
- Pre-built dashboards with metrics