Self-hosted monitoring service

nodbeat

A heartbeat for every service and device. A small, self-hosted dead-man's switch: cron jobs, servers, Raspberry Pis, and ESP32 sensors ping a secret Heartbeat URL, and nodbeat alerts on Telegram, a Webhook, or ntfy when one goes silent or fails.

nodbeat came from the same problem on both sides of my work: backups, scripts, home servers, and battery-powered sensors that fail quietly. Heartbeats are push-only, so nodbeat never connects to what it watches and devices behind NAT work without any port forwarding. It runs live at nodbeat.g2labs.pl (invite-only for now), and the code is public under the MIT licence.

nodbeat landing page with the heartbeat monitoring dashboard

One line to integrate

curl -fsS -m 10 --retry 3 https://nodbeat.g2labs.pl/hb/<token>

Product snapshot

Monitoring that works from behind NAT

Read the README
Push-only

works behind NAT

30 s – 7 d

interval or cron schedule

3 channels

Telegram, Webhook, ntfy

MIT

open source

Who it is for

For anything that should check in and sometimes doesn't

  • Developers and operators who need to know when a cron job, backup, or script stopped running
  • IoT and embedded builders whose devices sit behind NAT and can only make outgoing HTTP requests
  • Small teams and homelabs that want monitoring they can self-host instead of another SaaS account

How it works

Create a Monitor, send Heartbeats, get an Alert.

Each Monitor has a secret Heartbeat URL and an expected interval or schedule. When a Heartbeat is late, or a job reports a failure, nodbeat marks the Monitor Down and alerts you, then again when it is back.

TypeScriptFastifyMongoDBMonitoringIoTSelf-hostedOpen source

Product qualities

Small surface, real-world edge cases covered

Anything that can make an HTTP request

A cron job, a systemd timer, a Raspberry Pi, or an ESP32 with Wi-Fi: a GET or POST to the Heartbeat URL is all the integration it takes.

Interval or schedule

Expect a Heartbeat every few minutes, or at 02:00 on weekdays in a chosen time zone, with a grace period before a Monitor counts as Down.

Alerts that carry context

Down and recovery alerts include the last message a job sent, such as "battery 3.7V" or a failure reason, with reminders while a Monitor stays Down.

Features

Beyond "is it up?"

Feature

Start and Fail signals, so a failed run is Down at once and run time is measured

Feature

Max run time: a job that starts but hangs goes Down

Feature

Last message and change alerts, e.g. when "door=open" turns into "door=closed"

Feature

Reminders every 1, 6, or 24 hours while a Monitor stays Down

Feature

7-day uptime and incident history per Monitor, with paused time left out

Feature

JSON API and dashboard data for scripts and integrations

Why it matters for clients

The same engineering, applied to a whole product

Execution signal

A complete web product built and run end to end: backend, dashboard, alerting, API docs, Docker deployment, and a live instance

Execution signal

Designed around real device constraints: push-only heartbeats that work from behind NAT and from ESP32-class hardware

Execution signal

Developed in public with ADRs, a glossary, a changelog, and tests, the same discipline applied to client work