Compare runs
B is compared with the previous run of the same agent.
10 steps · 12.7s · ≈$0.050
9 steps · 10.6s · ≈$0.046
steps 10 → 9 · cost −9% · duration −17% · outcome succeeded
What changed between them
Same app version, agent version, prompt version and model.
Step by step
= same (7)≠ changed (1)+ only in B (0)− only in A (1)| A · run_90d74343 | B · run_1c9b8b0c | |
|---|---|---|
| = | #2 claude-sonnet-4-5 | #2 claude-sonnet-4-5 |
| = | #3 sentry.getIssue | #3 sentry.getIssue |
| = | #4 claude-sonnet-4-5 | #4 claude-sonnet-4-5 |
| = | #5 gitlab.searchCode | #5 gitlab.searchCode |
| = | #6 gitlab.readFile | #6 gitlab.readFile |
| = | #7 claude-sonnet-4-5 | #7 claude-sonnet-4-5 |
| ≠ | #8 jira.createIssue429 Too Many Requests | #8 jira.createIssueoutcome changed |
| − | #9 jira.createIssue | — |
| = | #10 slack.postMessage | #9 slack.postMessage |
System spans are left out; steps are aligned on the tool called (or LLM call), then compared on arguments and outcome.