Run explorer
A tool for inspecting AI agent runs. 5,000 runs, each with its model calls, tool calls and retrievals laid out on a timeline. The list renders only what is on screen, the panes resize, and everything works from the keyboard: / to search, ↑ ↓ or j k through runs, Tab into the steps.
Sample data, generated from a fixed seed. No real model or customer data. Read how it was designed.
/
- Runs
- 5,000
- Pass rate
- 79.2%
- p50 duration
- 3.49 s
- p95 duration
- 5.00 s
- Est. cost
- $177.80
| Duration | Runs |
|---|---|
| 0 ms to 497 ms | 0 |
| 497 ms to 994 ms | 1 |
| 994 ms to 1.49 s | 33 |
| 1.49 s to 1.99 s | 97 |
| 1.99 s to 2.48 s | 342 |
| 2.48 s to 2.98 s | 783 |
| 2.98 s to 3.48 s | 1217 |
| 3.48 s to 3.98 s | 1120 |
| 3.98 s to 4.47 s | 769 |
| 4.47 s to 4.97 s | 372 |
| 4.97 s to 5.47 s | 132 |
| 5.47 s to 5.96 s | 37 |
| 5.96 s to 6.46 s | 6 |
| 6.46 s to 6.96 s | 7 |
| 6.96 s to 7.45 s | 7 |
| 7.45 s to 7.95 s | 5 |
| 7.95 s to 8.45 s | 8 |
| 8.45 s to 8.95 s | 5 |
| 8.95 s to 9.44 s | 8 |
| 9.44 s to 9.94 s | 11 |
| 9.94 s to 10.4 s | 14 |
| 10.4 s to 10.9 s | 13 |
| 10.9 s to 11.4 s | 9 |
| 11.4 s to 11.9 s | 4 |
Passed, Support triagerun_00001 · Nov 9, 12:00 UTC3.96 shaiku
Passed, Contract searchrun_00002 · Nov 9, 11:59 UTC3.25 ssonnet
Passed, Refund eligibilityrun_00003 · Nov 9, 11:59 UTC2.55 ssonnet
Failed eval, Contract searchrun_00004 · Nov 9, 11:55 UTC3.66 sopus
Passed, Meeting schedulingrun_00005 · Nov 9, 11:52 UTC2.39 ssonnet
Passed, Code review summaryrun_00006 · Nov 9, 11:50 UTC4.42 ssonnet
Passed, Order statusrun_00007 · Nov 9, 11:49 UTC2.55 ssonnet
Passed, Meeting schedulingrun_00008 · Nov 9, 11:47 UTC3.58 ssonnet
Passed, Meeting schedulingrun_00009 · Nov 9, 11:46 UTC3.29 shaiku
Passed, Meeting schedulingrun_00010 · Nov 9, 11:43 UTC2.16 ssonnet
Passed, Meeting schedulingrun_00011 · Nov 9, 11:41 UTC3.98 shaiku
Failed eval, Invoice reconciliationrun_00012 · Nov 9, 11:39 UTC4.08 sopus
Passed, Order statusrun_00013 · Nov 9, 11:39 UTC2.22 ssonnet
Failed eval, Refund eligibilityrun_00014 · Nov 9, 11:36 UTC2.86 ssonnet
Support triage
Passedrun_00001“Triage ticket T-9902: billing question.”
- Score
- 0.82
- Duration
- 3.96 s
- Tokens
- 8,641
- Est. cost
- $0.01
- Model
- claude-haiku-5-5
- Started
- Nov 9, 12:00 UTC
Step
Support triageAgent
3.96 s
planLLM
507 ms
search_docsRetrieval
158 ms
classify_ticketTool
1.04 s
classify_ticket.llmLLM
1.00 s
search_kbTool
785 ms
assign_queueTool
69 ms
respondLLM
1.29 s
Support triage
AgentOK- Started
- +0 ms
- Duration
- 3.96 s
- Share of run
- 100.0%
Input
{
"prompt": "Triage ticket T-9902: billing question."
}Output
{
"score": 0.82,
"status": "passed"
}