Set up notifications

Notifications tell your team when an incident opens, changes, or resolves. A destination is where a message goes. A route decides which incident events go to which destinations.

Squasher sends notifications for incidents only. A new issue does not send a notification. To be alerted about errors, add a query monitor that counts error events. It opens an incident when the count passes your threshold.

What sends a notification

Incidents open from monitors and heartbeats, from alert sources such as Datadog or Grafana, and from people who declare them. Each change to an incident is one event:

Event	When Squasher sends it
`opened`	An incident opens, or someone reopens a resolved incident.
`acknowledged`	Someone, or an alert source, acknowledges the incident.
`status_changed`	A person changes the status, or an alert source updates the incident.
`resolved`	The incident resolves.
`public_update`	Someone posts a public update for the status page.
`comment_mentioned`	Someone mentions a teammate in a comment. See Comment mentions.

When nobody acknowledges an on-call page, the route that started the page also sends an escalation_exhausted notice to its other destinations, whatever events the route subscribes to. Scheduled incidents, such as planned maintenance, send no notifications.

Add a destination

Only owners, admins, and security admins can change notifications.

Open your project and select Settings in the sidebar, then the Notifications tab.
Under Add destination, choose the type in Channel and enter a Name.
Fill in the fields for the type. See the table below.
Select Add destination. Squasher adds the destination to the default route. If the project has no default route, Squasher creates one named “All incidents”.
Select Send test next to the destination, then check the result under Recent deliveries.

Channel	Fields	Notes
Telegram	Bot token, Chat ID or public username, Topic ID (optional)	See Telegram.
Microsoft Teams	Teams Workflows URL	See Microsoft Teams.
Email	Email recipients (comma or line separated)	1 to 20 addresses.
Webhook	Public HTTPS URL, Signing secret (optional)	HTTPS only. Private and local addresses are rejected.
PagerDuty	PagerDuty integration key	See PagerDuty.
Slack channel	Connected Slack workspace, Channel ID, Channel name	Needs a Slack workspace connected to the project. See Slack.
Slack war room	Connected Slack workspace, New channel prefix (optional)	Creates a private Slack channel for each incident.
On-call escalation (text and call)	Escalation policy	Pages people by text and phone call. See On-call escalation.

After you save a destination, Squasher hides its secrets. You cannot edit a destination: to change it, remove it with Remove and add it again. To let a coding agent set up a destination, select Copy prompt next to Add destination. The prompt has the project ID, your current destinations and routes, and the MCP and API calls. It tells the agent to ask you which destination type and which incidents you want, and to confirm each change with you first. A test message is sent as an opened event without an incident. Send test is not available for On-call escalation destinations; open a test incident instead. A PagerDuty test opens a real PagerDuty incident, so resolve it in PagerDuty afterward.

Choose which incidents go where

The Alert routes card shows each route as a rule: its conditions (Events, Source, Monitors, Severity, and the other filters it has), then Sends to with its destinations. Routes with filters come first; the default route is last. To change the destinations of a route, select Edit destinations, select or clear a destination, then select Save destinations. If you have destinations but no default route, select Create default route. Each route shows its ID and a Copy prompt button that asks a coding agent to explain the route or check its deliveries. Copy prompt next to Alert routes asks the agent to plan new routes with you. The default route, “All incidents”, sends the opened and resolved events of every incident. It runs only when no other route matches the event. A route can have up to 20 destinations.

Add a route with filters

The dashboard cannot create filtered routes yet. Use the Notifications API with an API key that has the notifications:write scope, squasher notifications routes create --input '<JSON>' in the CLI, or search for notifications in the Squasher MCP server. This route pages PagerDuty for sev-1 incidents only:

curl -X POST "https://api.squasher.ai/v1/projects/$SQUASHER_PROJECT_ID/notifications/routes" \
  -H "x-squasher-key: $SQUASHER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "SEV-1 pages",
    "conditions": {
      "severities": ["sev-1"],
      "events": ["opened", "acknowledged", "resolved"]
    },
    "destinationIds": ["<pagerduty-destination-id>"]
  }'

A route matches an event when every filter that you set matches. In a list, one match is enough.

Filter	Values
`scope.sourceTypes`	`monitor`, `heartbeat`, `manual`, `integration` (alert sources), or `finding` (anomalies).
`scope.sourceIds`	Source IDs, such as alert source (connector) IDs. Use it to send one alert source to its own route.
`scope.monitorIds`	Monitor IDs.
`scope.components`	Status page component names on the incident.
`conditions.severities`	Severity keys, such as `sev-1`.
`conditions.events`	The events in the table above. Without it, the route sends `opened` and `resolved`.
`conditions.attentionClasses`	`interrupt` or `digest`. Events are `digest` only while a monitor tuning override makes them quiet.
`behavior.includeAiSummary`	Slack only. `false` leaves out the root cause and the recommended action.

conditions.environments and conditions.tags are in the API, but incidents do not have these values yet. A route that uses them never matches.

What each destination receives

Each message shows the severity with the project’s label, for example SEV-1, and links to the incident.

Destination	Message
Slack channel	The title, severity, and status, then Impact, Root cause, and Recommended action. When a deploy came shortly before the incident, an `opened` message says how long after the deploy it started. Buttons open the incident. When AI triage finishes for a monitor incident, the result is posted in the message thread.
Slack war room	On `opened`, a new private channel named with the prefix (`inc-` by default) and a short incident ID. Later events post in that channel. Squasher archives the channel when the incident resolves. When you reopen the incident, Squasher unarchives the same channel and posts there.
Email	The subject `[SEV-1] <incident title>` and an Open incident button.
PagerDuty	An Events API v2 event: `trigger`, `acknowledge`, or `resolve`, with the incident ID as `dedup_key`. See PagerDuty.
Microsoft Teams	An Adaptive Card: the event, title, severity, and status, then Impact, Root cause, and Recommended action. Buttons open the incident and its monitor. See Microsoft Teams.
Telegram	A short text message: severity, title, event, and link.
On-call escalation	On `opened` only: starts the escalation policy. Squasher texts the person who is on call, calls if nobody acknowledges, and escalates level by level. Acknowledging or resolving the incident stops the page.
Webhook	A JSON `POST`. See below.

Webhook payload

{
  "event": "opened",
  "incident_id": "…",
  "project_id": "…",
  "title": "Checkout API is down",
  "severity": { "key": "sev-1", "label": "SEV-1", "rank": 1 },
  "sent_at": "2026-09-25T08:00:00.000Z"
}

When you set a signing secret, the x-squasher-signature header has the HMAC-SHA256 of the raw body, in lowercase hex, with the secret as the key. Compute the same value from the raw body and compare the two before you trust the request.

How delivery works

Squasher picks up new events every few seconds and sends each one to every destination on every matching route.
A destination gets each event of an incident once. When a delivery succeeds, Squasher never sends it again.
A failed delivery is tried again after 1 minute and after 5 more minutes. After the third attempt, Squasher stops. For Telegram and Microsoft Teams, Squasher also waits as long as the service asks.
Squasher does not try a Slack delivery again when the Slack setup is broken: the workspace is not connected, Slack rejects the token, or the bot cannot find or post in the channel. The destination shows Slack not connected when it has no connected workspace. When you connect Slack to the project again, its Slack destinations use the new connection.
Webhook requests time out after 10 seconds. Squasher does not follow redirects and does not retry a redirect.
Recent deliveries shows the last 12 deliveries, with Delivered, Failed, or Outcome unknown and the HTTP status. Outcome unknown means the request may have reached the destination, so Squasher does not send it again. The API returns up to 500.

Read your notification inbox

Select Notifications in the dashboard sidebar, or open the notification bell. Your inbox includes updates from all organizations and projects that you can access.

Update	Recipient
Incident mentions and assignment	The mentioned or assigned person.
Incident changes and response handoff	Incident participants or the selected responder.
On-call pages	The person selected by the escalation policy.
Investigation and fix results	The person who requested the work.
Fix merge, closure, deployment, or recurrence	The person who requested the fix.
Automation approval and daily budget limits	Project administrators.
Integration problems and recovery	Project administrators.
Final destination delivery failure	Project administrators.
Payment failure and subscription changes	Organization members who can manage billing.
Usage at 50, 80, and 100% of the included amount or budget	Owners and billing admins, also by email.
Invitations, member acceptance, and role changes	The invited person, inviter, or affected member.
Revoked API keys	The key owner when another person revokes the key.

Select an item to open the related work. Use Mark all as read to clear the unread count. Use Unread to show only unread items.

Limit repeated updates

Squasher stores one decision for each event and recipient. Retrying the same event does not create another notification. Your own actions do not notify you. Repeated updates for the same item are grouped while its inbox item is unread. Grouped updates do not send another email, except usage notifications: each usage threshold sends its own email. Mark the item as read to allow a later update after the one-hour cooldown. New notifications have hourly and daily limits. Additional updates increase a bounded summary instead of creating one notification per event. Use Preferences in the inbox to turn off optional categories. On-call pages, unanswered escalations, payment failures, and revoked-key notices remain enabled and have separate limits. Notifications from a person show their current profile photo. If the photo is missing or cannot load, their initials appear instead. System updates keep their category icon. Your inbox hides organization and project updates when you lose access. Billing updates require billing access.

Comment mentions

When you mention a teammate in an incident comment, the teammate can receive an inbox item and an email to their account address. Preferences, grouping, and notification limits apply to both. The author does not receive one. To also post mentions in a Slack channel, add a route that is not the default route, with comment_mentioned in conditions.events and a Slack channel destination. Mentions do not go to Telegram, email, webhook, or PagerDuty destinations. Comments and mentions are internal: they never appear on a status page.

Next steps

Set up on-call

Hand incidents to the person on call.

Connect alert sources

Open incidents from Datadog, Grafana, CloudWatch, and more.

Send alerts to Telegram

Set up a bot for a chat, group, or channel.

Notifications API

Manage destinations, routes, and deliveries over HTTP.

Get started

Guides

Set up notifications

What sends a notification

Add a destination