Operate
What happened to my app
One plain-English timeline per app, deploys, outages, changes, sleep and alerts, with the one button that fixes each.
Every app has a What happened tab: its recent history as plain sentences, newest first. "Version abc1234 failed to deploy: missing script: build. The previous version kept running." "Changed settings, by agent claude." "Stopped answering. Back up after 4 min."
Each line has the one action that makes sense for it:
- Roll back to this version on an older version that worked.
- Try again on a failed deploy, or See the logs.
- Undo on a change a person or an agent made.
- Wake it when the app is asleep.
- Read what happened on an outage.
How to use it: open an app and choose What happened. Pick how far back to look: a day, 3 days, a week or 30 days. The sentence at the top sums it up, like "In the last 3 days: 4 deploys (1 failed), down once, 2 changes to settings."
"What happened to my web app since yesterday?"
The endpoint
GET /v1/services/:id/story?hours=72hours is 1 to 720 (default 72). Anyone on the project can read it.
{
"service": { "id": "svc_…", "name": "web", "status": "running", "asleep": false },
"from": 1762400000000,
"to": 1762659200000,
"summary": "In the last 3 days: 3 deploys (1 failed), 1 change to settings.",
"items": [
{
"id": "deploy:dep_…",
"at": 1762659000000,
"kind": "deploy",
"title": "Version abc1234 failed to deploy",
"detail": "commit abc1234 · missing script: build · the previous version kept running",
"who": "a Git push",
"tone": "bad",
"actions": [
{ "kind": "retry", "label": "Try again", "serviceId": "svc_…", "commit": "abc1234…" },
{ "kind": "link", "label": "See the logs", "href": "/deploys/dep_…" }
]
}
]
}What goes in
kind | From | Notes |
|---|---|---|
deploy | the app's deploys | live (with how long it took), failed (with the error summary), cancelled, or in progress |
outage | incidents | "Stopped answering" and "Back up and answering after N min" |
change | the activity log, for changes aimed at this app | titled with the recorded intent (X-CLI-Intent) when there is one; agents are named by their key |
sleep | scheduled sleep and wake, and "Asleep right now" | see Sleep, schedules and instant wake |
alert | project events that name this app | cron failures, browser checks, log alerts and so on |
tone is good, bad, warn or neutral. Items are newest first, at most 150.
Actions
kind | What to call |
|---|---|
rollback | POST /v1/deploys/:deployId/rollback {"toDeployId": "…"} (offered on an older version with an image, when another version is live) |
retry | POST /v1/services/:serviceId/deploys {"commit": "…"} |
undo | POST /v1/flight/:entryId/undo (see /docs/flight-recorder) |
wake | POST /v1/services/:serviceId/wake |
link | a dashboard path |
Every action runs as the caller, so roles, approvals and spending caps apply as usual.