Every confirmed outage has a page at /incidents/<number>/. The View incident button in an email, Open incident in a chat message, the incident.url of a webhook and the rows of the Incidents list all lead there.
What it shows
The header gives the state and how long it has lasted, the reason with the HTTP code where there is one, the first city that saw it fail, «confirmed by N of M probes, K continents», and the time it was detected in the account's time zone.
The timeline lists, in order: the first failing round, the confirmation, every alert by channel with sent or failed and the error, who acknowledged and how, each update posted to a status page, the summary being written, and the recovery.
Working the incident
- Summary and severity. A short text the team writes, with a severity. On accounts where the AI features are turned on, Claude's reading of the probe evidence appears next to it; see AI explanations.
- Acknowledge. The same as the Acknowledge link in the alert: reminders stop, monitoring and the recovery alert carry on. See Acknowledging and escalation.
- Post an update. Writes an update to the status-page announcement linked to this incident, or opens one on a page you choose, and can publish it in the same step. Later updates go to the same announcement.
- Post-mortem. Once the incident is resolved, the Post-mortem tab holds the announcement's post-mortem, with the same draft and publish buttons as on the status page.
Copy for AI
Copy for AI puts the incident on the clipboard as Markdown: the header, the timeline, the quorum, the probes' results and the monitor's settings. The same text is served at /incidents/<number>.md to a signed-in browser or to an API key with Authorization: Bearer, so an agent can read it. It never contains request headers, secrets, the addresses alerts go to, or IP addresses, and it covers this account only. In the API, each incident carries url and markdown_url.
Last change: 6 November 2026.