Get alerts | ICPWatch

Subscribe to any public canister, or to a whole token: ledger, index, and every archive, including archives that get spawned later. You do not need to control the canister.

What to watch

--- Index i5e5b-eaaaa-aaaaq-aadpa-cai

Choose something else

Alert types

Low cyclesUnder 14 days until it freezes at the current burn.FrozenThe canister started rejecting calls.Approaching deletionFrozen with under 14 days of storage cost left.Burn-rate anomalyA spike against the canister's own rolling baseline.Controllers changedSomeone was added to or removed from the canister's controllers. The ICP version of "admin key changed".Code changedThe canister was upgraded, reinstalled, or uninstalled. Routine for active projects, so off by default.

Where to send it

EmailTelegramDiscordWebhook

Address

Email: we send one confirmation link first (email delivery is being set up, so subscriptions are stored and confirmed once it is live). Telegram: message the ICPWatch bot with /start to get your chat id (bot is being set up). Discord: paste a channel webhook URL. Webhook: we POST signed JSON, with the signing secret shown once after you subscribe.

One alert per incident, plus one note when it resolves. Every message has a one-click unsubscribe.

How alerts behave

One per incident. An incident opens once and stays open. You are not pinged again while the balance keeps falling.
Hysteresis. A low-cycles incident opens under 14 days of runway and only resolves after the runway is back above 21 days on two checks in a row, so a canister hovering at the line does not flap.
Own baseline. Burn anomalies compare the last six hours against the median and spread of that canister's own daily burn. A canister that is always busy does not fire. A quiet one that suddenly burns several times its norm does.
Resolved notice. When it recovers you get exactly one message saying so, with how long it was open.