Status pages and incidents
A public status page built from a workspace's flows and cron monitors, with incidents, maintenance, email subscribers and AI-drafted postmortems you review before they count.
A status page tells the people who use your service whether it works. In Yard it belongs to a workspace and is made of components, each built from that workspace's flows and cron monitors. The page shows each component's state now, 90 days of history, open incidents and scheduled maintenance. Anyone can read a published page without signing in.
Set up a page
On the Monitors page, open Alerts, status pages and incidents and press New status page: a title and an address, then Create status page. A new page is a draft. Its page has:
- Components: add one by name (Add component), tick the flows and cron monitors it is built from, reorder or remove it.
- Maintenance: Schedule maintenance with a title, details, start, end and components.
- Details: title, description and address, saved with Save details.
- Settings: Published (off, the address answers "not found"), Email subscriptions, and Incidents from failing monitors (Off, Open drafts or Open and publish).
- Subscribers, and View page and Unpublish or Publish page in the header. Its More menu has Incidents.
The public address
A page lives at <API address>/status/<slug>, for example
https://api.example.com/status/northport. This is its one canonical
address: subscriber emails, the Atom feed and the badge all link to it. The
page is plain HTML rendered by the API, so it reads without JavaScript, in
light and dark (following the visitor's system setting). The web app shows
the same pages at /status/<slug> from the same public data.
Next to the page:
| Address | What it is |
|---|---|
/status/<slug>/history |
Past incidents and maintenance |
/status/<slug>/incidents/<id> |
One incident with every update and, once published, its postmortem |
/v1/status/<slug>/summary.json |
A small JSON summary to poll: overall state, components, open incidents, next maintenance |
/v1/status/<slug>/feed.atom |
An Atom feed of incidents |
/v1/status/<slug>/badge.svg |
A badge with the overall state, for a README |
An unpublished page answers 404 everywhere, exactly like a page that does not exist.
What a visitor sees, and what they never see
A component's state comes from its monitors:
- A flow: its newest finished run. Passed is operational; failed, timed out or errored is down; inconclusive (the runner could not tell) or no run is no data. A last run older than about twice the flow's schedule counts as no data too.
- A cron monitor: healthy is operational; late, failed or stuck is down; new or paused is no data.
- Some monitors down and others up makes the component degraded; all of them down makes it down. Maintenance in progress shows maintenance.
The 90-day bars show each day from the checks that ran that day. Runs are kept as long as the workspace's run retention (30 days unless you change it), so older days show "No data".
Public answers carry the component's name and nothing else about its monitors: no URLs, step results, environment values, flow or cron names or ids. Draft incidents never appear, and a postmortem appears only once someone publishes it.
Incidents
Open an incident on a page with a title, an impact and a first update. Each update sets the incident's status (investigating, identified, monitoring, resolved). A published incident is on the page at once and emailed to subscribers; a draft stays private until you publish it.
Auto-incidents. A page's auto-incident setting is off, draft (the default) or publish. When a monitor of one of its components starts alerting, the page gets one incident. Another monitor of the same page failing within 30 minutes joins that incident instead of opening a second one. Its title and updates name only components. When every monitor of the incident has recovered, Yard proposes resolving it; a person confirms.
Maintenance
Schedule a window with a start, an end and the components it affects. While it runs, those components show "maintenance". Subscribers get one email when it starts and one when it ends.
Email subscribers
With subscriptions turned on, visitors can subscribe from the page. They get
a confirmation email first and nothing else until they confirm. Confirmed
subscribers get an email for every update of a published incident and for
the start and end of maintenance. Every email has an unsubscribe link and a
one-click List-Unsubscribe header. Subscribing is rate limited per address
and per page.
AI-drafted briefs and postmortems
On an incident's page, Draft brief or Draft postmortem asks your own agent account (Claude Code, as for other one-shot questions) to write a draft:
- Yard collects the evidence from the incident's time range: runs and failed steps (status, code, error and assertion results; never response bodies), cron pings and executions, alerts and the incident's updates. Secrets are masked on your runner, including every secret value of the workspace's monitor environments, before anything is stored or sent.
- The agent answers with observations and hypotheses, each citing evidence. A claim without a valid citation is dropped (the draft says how many).
- You read the draft next to its evidence and apply the fields you choose to the postmortem. Nothing reaches the postmortem before that, and the postmortem stays unpublished until you publish it.
The postmortem can be copied as Markdown with the incident's timeline.
Who can do what
Everyone who can see the workspace can read its pages, incidents and
subscribers; editors create and change them. The routes are in the API
reference under status.