Analytics
Aggregate dashboards for cost, latency, errors, tools, and detection trends across your traces.
Where the Traces view is for inspecting one run, Analytics is for seeing the whole picture across many. It aggregates your traces into a configurable dashboard so you can track cost, latency, error rates, tool usage, and detection trends over time — and spot regressions before they become incidents.
Widgets
The dashboard is built from widgets you can show, hide, and reorder:
- KPI strip — headline metrics: total traces, error rate, cost, token usage
- Cost breakdown — spend by model, workflow, or time, so you can see where tokens are going
- Time breakdown — where wall-clock time is spent across spans and tools
- Error analysis — error rates and recurring failure patterns, with similar errors grouped together
- Tool performance — latency and call counts per tool/function
- Detection trends — how often each detection is firing over time
- Arc patterns — common paths traces take through your agent, and where they branch toward success or failure
Filtering
A filter bar scopes the entire dashboard — by time range, workflow, model, tool, or detection — so you can answer questions like "what did the customer-support workflow cost last week?" or "which tool is driving our p95 latency?" without leaving the page.
Relationship to traces
Every number on the Analytics dashboard is computed from the same span data you inspect in Traces. When a widget surfaces something — a cost spike, an error cluster — you can drill from the aggregate straight into the individual traces behind it.