Alerts

Channels

ChannelPlans
EmailFree, Starter, Pro
SlackStarter, Pro
Generic webhookPro

Email ships over SMTP via Postmark. Slack alerts post through CertWatch's own Slack app to the channel you choose after Add to Slack on the Alerting page; the bot token is stored encrypted. Generic webhooks are sent directly from CertWatch's servers to the URL you provide — handling past that point is whatever you've connected it to (see Trust & security).

Email alerts go only to an address you've verified, using the link CertWatch emails you. The Send test email button on the Alerting page works the same way: it works once your alert email is verified, and the app's banner can send the verification link again. Test emails are limited to a few per hour.

Alert channels

The Alerting page has an Alert channels card above its tabs, one setting for the whole account, covering every endpoint and domain: a row for each channel, with a chip for each severity, Late, Urgent, Down, and Unreachable, filled when that severity's alerts go to the channel. Select a chip to turn it on or off. By default every severity goes to every channel. The card only narrows what your plan allows: a channel your plan does not include, one that is not set up, or an alert email that is not verified is greyed out and gets nothing, and the info icon beside its name says why when you point at it, and links to Billing when your plan is the reason. A severity already on for it stays on, so it sends again once the channel can. When none of the channels on for a severity can send, its alerts go to email instead, if the alert email is verified, and a severity with every channel off sends nothing; the card warns about both. A change applies from the next alert and is recorded on History. A domain sends each status's alert once, with no reminder, so a channel turned on while a status lasts gets that alert, as first sent, at the domain's next check. An outage on an endpoint whose certificate is Late is an Unreachable alert, so it goes to the channels on for Unreachable. A Recovered goes to every channel that got an alert about the problem, unless the endpoint is muted.

Slack mention

The Alerting page's Slack tab has a Slack mention dropdown: No mention, @here or @channel. An endpoint's Urgent and Down alerts start with it in Slack. Other alerts, email, and the webhook carry no mention.

Alert types

Every alert has one of five types, the status it announces. The email subject, the webhook's type, and Alert history all lead with it. A Slack post's first line names it and the monitor, as in "Alert: Down - Checkout (checkout.example.com)", after the account's Slack mention, if it has one, on an endpoint's Urgent or Down post. A Slack post has a bar in its status colour and a View monitor button in place of the links other channels carry, which opens the monitor card's Overview tab.

A certificate that renews on schedule sends nothing at any point of its life, and a new certificate, issuer, or key is never an alert (see a new certificate).

When an endpoint alerts

An endpoint's alerts follow its status. CertWatch sends an alert when the status becomes Late, Urgent, or Down, sends it again while the status lasts (a reminder), and sends one Recovered when the endpoint is Healthy again, unless the endpoint is muted. Removing the monitor ends the problem with no Recovered. How often it reminds depends on the status, not on the renewal mode:

StatusReminder
LateEvery 7 days
UrgentEvery 24 hours
DownEvery 24 hours
UnreachableEvery 24 hours

A status CertWatch has not yet sent you sends its alert at once, from whichever check finds it: a scheduled check, a Check now, or the check that adding or resuming the endpoint, or changing its renewal setting, runs. That is a worse status, or a drop to one never sent, such as Down falling to an Urgent you were never told of. A drop to a status already sent sends nothing, and its reminders then follow its own interval; a return to the worse status after that is sent at once again. When a finding such as a weak key made the endpoint Urgent first, the renewal date's own Urgent point is still sent once, and when a finding made it Urgent or Down first, the certificate's expiry is still sent once. Reminders go out only with a scheduled check, with no limit on how many. Each one says what changed since the last alert: findings that came or went, a different certificate, a moved renewal date, the time left.

A finding that depends on what a check saw, such as a chain that does not verify, alerts only once it is confirmed (see Findings). An endpoint added when its certificate is already past its Late point gets the alert for its status then, which says so and points to the monitor card's Renewal tab to set the renewal mode.

Failed checks. A failed check (connection error, DNS failure, timeout) leaves the endpoint at its last status in the app and sends no Unreachable alert, because transient network failures are common; the endpoint says since when its checks have failed, and CertWatch checks it again about two minutes later (when the failed check was not itself such a check, nor one CertWatch could not finish on its own side). The exception is the certificate CertWatch last saw reaching its Late or Urgent point or expiring: that alert goes out at once, even from a failed check, as described below. If that follow-up check fails too, the endpoint is Unreachable and CertWatch sends an Unreachable alert naming the addresses that failed and the certificate it last saw, and repeats it every 24 hours while the checks fail. With no follow-up check, or one that is missed, the next failed check confirms the outage instead. Statuses rank Down, Urgent, Unreachable, Late, so an endpoint whose Late alert is open becomes Unreachable then, while an Urgent or Down alert that is already open keeps its status and its reminders through failed checks, measured on the certificate last seen, and that certificate reaching its Urgent point or expiring sends that alert at once. An endpoint CertWatch has never read is Unreachable from its first failed check, and its alert still waits for the follow-up check. An endpoint whose addresses CertWatch refuses to connect to (private or internal addresses) is Unreachable but never sends an Unreachable alert. A check CertWatch could not finish on its own side ("CertWatch could not finish the check": no DNS answer, no time or IPv6 route to try an address, or an issuer fetch that timed out) never counts toward Unreachable, gets no follow-up check, and sends nothing of its own; only the certificate last seen reaching its Late or Urgent point or expiring still alerts.

Recovered. One Recovered ends each problem that sent an alert, however many it sent, unless the endpoint is muted when the problem ends; a problem that sent none (every channel off or unusable for its severity) ends with none. It says when the problem was first seen, what the check that ended it read, and what the last alert found. A check that reads the certificate again at Urgent or Down after an Unreachable alert sends that status's alert instead, and one that reads it at Late sends one Late alert, even while the alert is acknowledged. Its email and Slack post are headed "Reachable again · still Late", and on every channel it opens with the check that read the certificate and when the checks started failing, then says what it found. Removing the monitor ends its problem with no Recovered: the alerts were about a monitor that is gone.

A muted endpoint (see Monitoring behavior) is checked the same as any other but sends no Late, Urgent, Down, or Unreachable alert while muted; the first check after the mute ends sends the one alert that came due, once: "Reachable again · still Late" when the problem's last alert was Unreachable and that check reads the certificate at Late, otherwise a new status, otherwise the reminder. A mute holds the Recovered too: a problem that ends during the mute sends no Recovered, then or when the mute ends.

When a domain alerts

A domain registration sends one alert for each status it reaches: Late at 30 days or fewer, Urgent at 7 days or fewer, and Down once it expires. None is sent again while the status lasts. CertWatch cannot see whether a domain renews automatically at its registrar, and some registrars renew only near or on the expiry day, so the Late alert asks you to check that auto-renew is still on and the payment method is valid. A domain added when its registration is already past a point gets the alert for its status then, not one for each point it passed. Renewing the registration sends Recovered. A domain whose registration lookups keep failing for 24 hours sends one Unreachable alert, and a Recovered once a lookup reads it again and it is Healthy. A domain on a TLD with no registration data never alerts.

Acknowledge

In Slack, an endpoint's Late, Urgent, Down, and Unreachable alerts have an Acknowledge button that anyone in the connected channel can use, with no CertWatch sign-in. It acts at once: the bar takes the acknowledged colour, the post shows who acked ("Acked by Sam (Slack)"), and the button becomes Unacknowledge. A newer message of the alert takes the buttons off the older post, and a click on a post that has moved on changes nothing.

In email and the webhook's text, those alerts end with an Acknowledge link; a Recovered has none. It opens the endpoint's monitor card on its Settings tab, after you sign in, to confirm: the link changes nothing by itself, so a mail scanner or a forwarded email cannot act on it. The same button is in the monitor card's header, on the endpoint's row in Monitors, in the Dashboard's Triggered pane while the endpoint is Triggered, and on the endpoint's row in the Triggered list.

Acknowledge answers the endpoint's open alert, once one has been sent: its reminders, at any status, stop until the alert ends. Any alert that is not a reminder (a status not yet sent, the renewal date's Urgent point or the certificate's expiry, or "Reachable again") still goes at once and ends it. An acknowledged Down alert whose expiry has been sent and that is never fixed stays silent until the certificate is renewed or the endpoint recovers. Withdrawing an acknowledgement, with Undo on the endpoint's monitor card or Monitors row, on its row in the Dashboard's Triggered list while it is Triggered, with Unacknowledge on the Slack post, or through the API, ends it at once: its reminders count from the last alert again, so one it held goes at the next check unless a mute holds it.

It does not hold a Recovered, and it is recorded on the endpoint's History. A mute holds every alert while it lasts, for a set time or until you unmute, and holds the Recovered too; an acknowledgement holds only until the alert changes. Domains have neither.

A real alert

The wording CertWatch sends for an Urgent alert, with a plausible label, hostname, and dates filled in. The email's subject:

Subject: Urgent: api.example.com — certificate not renewed, expiry is close

The same alert's text, which the webhook sends as its summary:

CertWatch Urgent: Production API (api.example.com): no renewal seen, and expiry is close.
The certificate expires on 2026-10-01 12:00 UTC (7 days left).
Late from 2026-09-03 12:00 UTC, based on the CA's suggested renewal window (ARI), which this endpoint's client follows; Urgent from 2026-09-17 12:00 UTC.
How to fix: 1. Check your renewal client's log to see whether it ran and succeeded. 2. Compare the serial served (0A1B2C) with the certificate on disk: a renewal that was never deployed looks the same from outside. 3. Check that the server reloaded after the renewal. 4. Check the CA's status page for an issuance or validation outage.
Renewal dates and settings: https://app.certwatch.sh/monitors/ep_123/settings
Acknowledge: https://app.certwatch.sh/monitors/ep_123/settings?confirm=acknowledge

The email lays out the same pieces: what was found, the expiry, the renewal dates (Late from and Urgent from) and where they come from, what changed since the last alert, the How to fix steps as a numbered list, and buttons for the monitor card's Settings tab and Acknowledge. While CertWatch is still learning when an endpoint renews, a Late alert about an overdue renewal opens its How to fix with "This endpoint is set to automatic renewal. CertWatch is still learning when it renews, so this may be early for your client." above the same steps. When a finding other than the renewal dates or a partial rollout set the alert's status, such as a weak key or a hostname mismatch, the email, the Slack post and the text leave the renewal dates out and keep the expiry. An Unreachable alert always keeps them, and the webhook still sends both dates.

The Slack post is shorter. Its first line reads "Alert: Urgent - Production API (api.example.com)", after your Slack mention if you set one. It shows the heading "Certificate not renewed, expiry is close", anything the heading and the dates cannot say, the expiry and renewal dates as labelled fields, where those dates come from in small grey text, and Acknowledge and View monitor buttons. It leaves out the How to fix steps, which are in the Why box on the monitor card's Overview tab, one click away on View monitor. An alert without these pieces, such as a Recovered or a domain's alert, shows its text without the link lines.

Alert history

Alert history lists every alert from the last 90 days, newest first, with what happened to it on each channel. Open it with the Alert history button on the Alerting page, and a View alert link on a monitor's History opens one alert's page. Each row shows the alert's message, when it was created, the monitor it is about, its type (Late, Urgent, Down, Unreachable, or Recovered), and a badge per channel such as "Email: Accepted". Older and Newer page through the list.

A test send, from the Alerting page or from the Dashboard's Alert delivery tile, is not an alert: it reports its result on that page or on the channel's row, and is not recorded in Alert history or in a monitor's History.

Alert details

Selecting an alert opens its page: the type, the message, a link to the monitor, and a card for each channel it was sent on. A channel's status is one of:

StatusMeaning
SendingThe first attempt is in progress.
RetryingAn attempt failed and the next automatic retry is scheduled, at the time shown. While a retry you started is in flight it also reads Retrying, with no scheduled time.
AcceptedThe email server, Slack, or your webhook endpoint accepted the message. This does not mean anyone has read it.
FailedThe attempts ran out, or the destination rejected the message in a way a retry would not fix. Correct the destination, then retry.
No longer applicableIt was not delivered, and the alert no longer applies, so it will not be retried.

Each card lists the attempts with their start time, outcome, and detail, such as "Could not reach the destination." A channel that was not configured or not usable when the alert fired is listed in a line under the cards as not sent. When an alert is no longer applicable, a note at the top says why: the status it reported ended, the status it reported as recovered came back, or the monitor was removed or paused.

A failed channel has a Retry button while the alert still applies. A retry uses the channel's current settings, so after you correct a destination, retry to send it there. A retry sends the alert's original message again on that one channel only; channels that accepted it are never sent again, and the attempt is marked "(manual)". An alert keeps the wording it had when it was created, so a retry of an older alert reads as the older text. An alert that does not exist, is older than 90 days, or belongs to another account shows a not-found message. For the payload a webhook receives, see Webhook payload on the API page.

Alerts, Notifications, and Notices

CertWatch uses a different word for each way it tells you something, so the screens and emails read the same way:

WordWhat it is
AlertA message about a monitor, sent to your alert channels: Late, Urgent, Down, Unreachable, or Recovered. Each one is recorded in Alert history and in the monitor's History.
NotificationAn item in the bell in the app's header: a to-do worked out from your account as it is now, never stored, and not dismissible (see the bell).
NoticeAn email about usage or billing, such as reaching your plan's endpoint limit. A Notice is never an item in the app. Usage emails go to every account whose alert email is verified.
Delivery noticeA message that one of your alert channels is failing, or is accepting alerts again, sent to a different usable destination from the failing one: another channel, or your account's sign-in email when the alert email is the one failing. It is not an alert and is not recorded in Alert history.
Alert channelA place alerts go: email, Slack, or webhook.
Alert emailThe address that receives email alerts. Only a verified address receives them.

We alert on our own monitoring going dark

A dedicated canary check is pointed at a permanently-expired public test certificate, so it's designed to fail on every run. We don't alert on that expected failure — we alert if that check's result ever stops reporting entirely, which is how we'd notice our own monitoring pipeline had gone dark rather than trusting "no alerts" to always mean "everything's fine."