Latest updateInvestigating22:26 UTC ()
p95 latency on the public API is above 500 ms. Requests are completing, more slowly than usual.
- Identified22:52
A slow query on the usage endpoint is holding connections. The fix is rolling out now.
ProductStatus pages
It changes when your metrics do, shows each customer their own commitments, and lives on your domain. Nobody has to remember to update it.
Current status
Latest updateInvestigating22:26 UTC ()
p95 latency on the public API is above 500 ms. Requests are completing, more slowly than usual.
A slow query on the usage endpoint is holding connections. The fix is rolling out now.
At a glance
Uptime, 90 days
99.76%Services operational
3of 4Active incidents
1API latency under 500 ms
API
API availability
API
Webhook delivery
Webhooks
Dashboard availability
Dashboard
Webhook deliveries are running up to ten minutes late.
Workers are back to full capacity. Watching the queue drain.
The queue has drained and every delayed event was delivered.
Status
Each metric is measured against two thresholds, and a value exactly on one does not count. The page follows the measured status, not someone's memory of updating it.
The reading
API · Request latency p95
642ms
The rule
Degraded
Neither rule matches: between the two thresholds is degraded.
One bad sample does not flip the page. By default a worse verdict has to hold for 3 minutes before the page follows, a better one for 5. Both are set per metric.
The page
Incidents
A minor incident reads degraded even while every metric is healthy, and a real outage stays an outage whatever was posted.
Customers
Sign customers in through your app. The same page shows each one the targets in their contract and the incidents that affect them, and nothing about anyone else.
Your dedicated export worker is behind. Exports finish, up to 15 minutes late.
10:55 PM UTC
Access
One mode per page. The badge and widget follow it, so a private page never leaks through an embed.
Reliability
Uptime is counted from recorded readings, day by day. Publish your SLO targets with the budget left, and give customers a Reliability Center with monthly attainment.
Uptime
Uptime counts the time a check was actually unhealthy, out of the time it reported a status, so a three-minute outage costs three minutes. Degraded time was served, so it counts as up, and time without data is left out instead of counted as up. The bars mark an hour down once it has two unhealthy readings.
Service levels
Choose which SLOs the page shows. Each row carries its target, its window and how much error budget remains.
Incidents
When a metric breaks, Observer can draft the incident for you. Check it, publish it in one step, and the page and its subscribers hear about it together.
Draft
A metric that turns unhealthy can open a draft quoting the reading against its threshold. Drafts never reach the page until you publish.
Publish
The incident lands under the headline, subscribers are emailed and your alert channels are notified.
Resolve
Every update and the resolution reach the same subscribers, so nobody has to check back.
Subscribers
Customers who can see the status don't open a ticket to ask. They subscribe from the page, and your app and docs show it where they already are.
Subscribe, from the page
Subscribe to updates
Get incident and maintenance updates by email, Slack, Teams or webhook. Unsubscribe at any time.
Confirmed by email. Each subscriber picks every change or a daily digest, and can leave from any message.
Status, in your docs
Webhooks
Degraded performance
Acme
The badge and the widget are served from the page's own address and follow its access mode.
Incidents scoped to a customer email only that customer's subscribers.
Slack, Teams, webhook
Visitors can send updates to a Slack or Microsoft Teams channel, or to a signed webhook, for every service or only the ones they pick. Each counts as a subscriber.
Feeds and JSON
For readers, calendar apps, chat integrations and your own dashboards.
Widget and badge
The badge can show 30-day uptime instead. Its color still follows the current status, so a degraded page never hides behind a green number.
<iframe src="https://status.example.com/embed" width="360" height="240" style="border:0"></iframe>
[](https://status.example.com)
Brand
Three layouts, eight theme presets with editable tokens, ten languages including Arabic and Chinese, your own address with the certificate handled, and custom CSS on Enterprise. Nothing reaches visitors until you save.
Minimal Dark
Minimal Light
Corporate Blue
Monochrome
Pastel
Serif Newspaper
Terminal Green
Vibrant Card

The page builder: pick a layout and sections, preview against your real metrics at desktop and phone width.
The one record
What the console shows
Domains per plan
Theme
A color that would fail WCAG AA is lifted until it passes, keeping its hue. Blueprint keeps its own ink palette and takes your logo and accent.
Proof
status.use.observer is an Observer page, reading Observer's own metrics. Yours can be public in about ten minutes.


Agent
One container next to your data. It only dials out; nothing connects in.
Metric
A query and two thresholds per metric, compared with strict operators.
Page
Pick the metrics, a layout and a theme, then publish. The page follows the signal from then on.
Keep reading