What it does
Assembles the timeline during an incident, while people are still too busy to write anything down. It pulls the error signal from Sentry, the paging history from PagerDuty, and what shipped in the hours before from GitHub, then lays them on one clock so the question 'what changed just before this started' has an answer instead of a guess. The correlation is offered, never asserted. A deploy fourteen minutes before the first error is a lead worth checking; the agent says exactly that, with both timestamps, rather than declaring a cause. Naming the wrong cause early is how an incident review goes wrong, because everyone stops looking. Afterwards it drafts the postmortem from what actually happened — first error, first page, first human acknowledgement, what shipped, when it was mitigated — with the gaps left visible. A timeline with three honest holes is more useful than a smooth narrative someone filled in from memory a week later. It is read-only on every production system. It cannot resolve an incident, silence an alert, roll back a deploy, or close an issue. During an incident the worst thing an automated tool can do is act.
Example prompts
- Checkout started erroring around 14:20. Build me the timeline.
- What shipped to production in the six hours before this error first appeared?
- How long was it between the first error and the first page?
Before you connect
The credentials this agent will ask you for — the full setup is on the Setup tab.
Needs 3 credentials (2 required) to connect. See setup
