More Updates
ChangelogInsights dashboard for spend and session health
RSS

Insights dashboard for spend and session health

·Console

A new Insights page in console charts spend over time, breaks it down by model and session, compares against the previous period, and surfaces prioritized observations about cost outliers and failed runs.

Console has a new Insights page that turns your session history into a read-out of spend and run health for a chosen window.

A stat strip reports total spend, tokens used, agent time, tool calls, and success rate, each with a delta against the previous same-length period. Below it, a spend-over-time chart plots daily cost — click a day to pin its top sessions and jump into them. Alongside are breakdowns of spend by model and your longest or most expensive sessions.

Insights
Total Spend
$1,284.52
18%
$6.00 avg
Tokens Used
1.94B
12%
1.93B in / 12.6M out
Agent Time
20d 6h
9%
2h 16m avg
Tool Calls
18.43K
21%
86 avg
Success Rate
92%
2.4 pts
197/214 complete
Spend over time
Daily breakdown - last 30 days
Live demo — hover the chart for a day's spend; click a day to pin its top sessions

The suggestions column flags what’s worth a look: a run that cost several times your median, spend on sessions that failed or were abandoned, and a high abandonment rate. Each links to the session it’s about; when nothing crosses a threshold, it shows a calm all-clear.

Suggestions
  • One session cost 468× your median
    $160.24 in a single run this period.
  • $2.73 on runs that didn't finish
    6 sessions failed or abandoned. Worth a look at what went wrong.
Spend by model
Total cost per model
Claude Fable 5
$612.3884 sess
Claude Opus 4.8
$398.2041 sess
Claude Sonnet 5
$182.1147 sess
GPT-5.2 Codex
$77.6323 sess
Claude Haiku 4.5
$14.2012 sess
unknown
$0.007 sess
Longest sessions
High turn count may indicate looping or inefficiency
Live demo — hover a session tag for details; flip the sessions card to Most expensive

The period selector is shared with the Sessions overview. Open Insights from the sidebar to try it.

Start with paper

Turn every session into knowledge at team scale.

Get started