For DevOps teams

Outside-in monitoring your pipeline can drive

AlertSleep checks your public endpoints, ports, certificates and scheduled jobs from outside your network, confirms failures before it pages anyone, and takes orders from your CI over a plain REST API. Nothing to install on your servers.

  • Checks every 60 seconds
  • Maintenance windows from CI
  • SMS and phone calls included
Mute alerts for a deploy, in any CI that runs curl
# .github/workflows/deploy.yml
- name: Mute AlertSleep during the deploy
  env:
    ALERTSLEEP_TOKEN: ${{ secrets.ALERTSLEEP_TOKEN }}
    MONITOR_ID: ${{ vars.ALERTSLEEP_MONITOR_ID }}
  run: |
    curl -fsS -X POST https://api.alertsleep.com/public/v1/maintenance-windows \
      -H "AuthToken: $ALERTSLEEP_TOKEN" \
      -H "Content-Type: application/json" \
      -d "{\"title\":\"Deploy ${GITHUB_SHA::7}\",
           \"starts_at\":\"$(date -u +%FT%TZ)\",
           \"ends_at\":\"$(date -u -d '+20 min' +%FT%TZ)\",
           \"monitor_ids\":[\"$MONITOR_ID\"]}"

Drive it from the pipeline

No plugin to install and none to keep up to date: every step is one HTTP request.

Watch a new service

After the first deploy of a service, create its monitor in the same job and keep the returned id for later steps.

# after a new service is deployed: watch its health endpoint
curl -fsS -X POST https://api.alertsleep.com/public/v1/monitors \
  -H "AuthToken: $ALERTSLEEP_TOKEN" \
  -H "Content-Type: application/json" \
  -d "{\"friendly_name\":\"$SERVICE (prod)\",
       \"url\":\"https://$SERVICE.example.com/healthz\",
       \"type\":\"http\",\"interval_seconds\":60}" \
  | jq -r .monitor.id

Take a server out of service

Pause its monitors while it is drained or rebuilt, and resume them when it is back. A paused monitor sends no alerts.

# taking a server out of service: stop checking it
curl -fsS -X PUT https://api.alertsleep.com/public/v1/monitors/$MONITOR_ID \
  -H "AuthToken: $ALERTSLEEP_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"status": "paused"}'

# back in service
curl -fsS -X PUT https://api.alertsleep.com/public/v1/monitors/$MONITOR_ID \
  -H "AuthToken: $ALERTSLEEP_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"status": "up"}'

Prove the backup ran

Report the start, the success or the exit code. A run that never starts, never ends or fails sends an e-mail.

#!/usr/bin/env bash
# nightly backup: report the start, the success or the exit code
set -o pipefail
PING=https://api.alertsleep.com/v1/ping/YOUR-TOKEN
curl -fsS "$PING/start"
if pg_dump "$DATABASE_URL" | gzip > /backups/db.sql.gz; then
  curl -fsS "$PING"
else
  curl -fsS "$PING/fail" -d exit_code=$? -d message="pg_dump failed"
fi

Public API at /public/v1, an API token in the AuthToken header, up to 500 requests a minute per token. A maintenance window lasts at most 24 hours, so a pipeline that dies halfway cannot mute your monitors for good. More examples for developers

Fewer false alarms at 3 a.m.

An alert that turns out to be nothing teaches people to ignore the next one.

Confirmed from a second location

When an HTTP or keyword check fails, the same request is repeated at once from a second, independent location. If it gets through there, nobody is paged.

Consecutive failures

Choose how many failed checks in a row, from 1 to 10, open an incident when no second opinion is available, for example on a port monitor.

Maintenance windows

Planned work raises no alerts and does not lower your uptime. Open a window in the dashboard or from the pipeline, and end it early when you are done.

Your own timeout

Between 5 and 60 seconds per monitor, so a slow but healthy endpoint stays green and a hung one goes red.

Channels per monitor

Production can call your phone while staging only posts to a chat channel.

Told when it is over

A recovery notice on the first good check, and the incident with its cause in the history.

Where alerts go

Every channel enabled on a monitor is alerted at the same moment.

E-mail SMS Phone call Slack Microsoft Teams Discord Telegram WhatsApp Webhook
All integrations
  • The alert says what failed: the status code, the timeout, the DNS, connection or TLS error, or the missing keyword.
  • The webhook sends JSON on monitor.down, monitor.up and certificate events, signed with HMAC-SHA256 when you set a secret.
  • SMS and phone calls come with the plan, from $59 a year, as a monthly allowance with no credit packs to buy.
  • Phone calls and SMS can be read in English, German, French, Spanish, Polish or Czech.

What AlertSleep is not

So you know before you sign up.

  • It checks from outside, so it reaches only public addresses. Services on a private network or behind a VPN cannot be monitored, and there is no agent to install.
  • It records availability and response time, not metrics: no CPU or memory, no query performance, connection pools or replication lag, no logs or traces.
  • It does not discover anything. There is no Kubernetes integration; each endpoint or port is a monitor you create, in the dashboard or through the API.
  • There are no CI plugins, Terraform provider or YAML configuration. curl in a pipeline step does the same job.
  • There is no PagerDuty or Opsgenie integration and no on-call rotation. The generic webhook can feed a tool that accepts incoming webhooks.
  • A port check is a TCP connection and nothing more; UDP is not supported.

Frequently Asked Questions

How does AlertSleep fit into a CI/CD pipeline?

Through the REST API, with curl or any HTTP client in a pipeline step. Open a maintenance window before a deploy and end it afterwards, create a monitor for a new service, or pause the monitors of a server you take out of service. There are no plugins to install.

How do I avoid alerts during a deploy?

Open a maintenance window for the affected monitors, from the dashboard or with a POST to /public/v1/maintenance-windows, and end it when the deploy is done. Checks keep running but no alerts go out, and the window does not lower your uptime. A window lasts at most 24 hours.

Can AlertSleep monitor Kubernetes?

It can monitor what Kubernetes exposes publicly: ingress hosts, health endpoints and TCP ports on public load balancers. It does not talk to the cluster, discover services or see pods, so each endpoint is a monitor you create, by hand or through the API from your deploy job.

Can I monitor databases and internal services?

Only if they are reachable from the internet. A port monitor opens a TCP connection to a public address and reports whether it was accepted. Private addresses are refused, and query performance or replication lag are not measured.

Does AlertSleep integrate with PagerDuty or Opsgenie?

Not directly. Alerts go to e-mail, SMS, phone calls, Slack, Microsoft Teams, Discord, Telegram, WhatsApp and a generic signed webhook. To use an on-call tool, point the webhook at an integration of that tool that accepts incoming webhooks.

How quickly will I know about an outage?

Checks run every 60 seconds. A failed HTTP or keyword check is repeated at once from a second location, and when both agree the alert goes out immediately. Port monitors wait for the number of consecutive failures you set.

Is there a Terraform provider or YAML configuration?

No. Monitors are created in the dashboard or through the REST API, which a script or a pipeline step can call to keep monitors in line with what you deploy.

Can I try it for free?

Yes. Both plans start with a 14-day trial with the API, maintenance windows and SMS alerts included, and no card required. After the trial, plans start at $59 a year.

Put your first endpoint under watch

Add it in the dashboard or with one API call, then mute it from your next deploy. Try it free for 14 days, no card required.