Alerts you can act on, and that clear themselves
Disk pressure, memory pressure, restart loops, failing health checks, expiring certificates, failed deployments and backups, and the traffic rules — delivered in the portal, by email, by signed webhook, or to Telegram. Every rule both opens and resolves, so the list stays worth reading.
What you get
One alert per condition
Alerts deduplicate on what they are about, so a check running every fifteen seconds raises one alert, not hundreds.
They resolve themselves
When the condition stops being true, the alert closes. A list of things that stopped mattering days ago teaches you to ignore it.
Your channels, your kinds
Email, signed webhooks, and Telegram, each subscribed to its own set of alert kinds — or to every kind, including ones added later.
Quiet when it is good news
On Telegram a resolution is sent silently: your phone wakes for a filling disk, not for the same disk being fine again.
In detail
- Server: disk above 85%, memory above 90% for five minutes, and container restart loops.
- Applications: failing health checks, failed deployments, and certificates expiring within fourteen days.
- Services: failed backups.
- Traffic: high 5xx rate, degraded p95 latency, traffic drops and spikes, AI crawl surges, spoofed agents, and analytics gaps.
- Memory pressure has a sustain window, because a build spikes memory for thirty seconds by design.
- Webhook deliveries carry a signature and timestamp header; a Telegram channel can be tested before the first real alert.
Works well with
Server monitoring
CPU, memory, disk, network, and every container — including steal time.
Request analytics
Every request to every app — read from the proxy, with no script to install.
Managed databases
One-click PostgreSQL, Redis, and MinIO with scheduled backups.
Your server is ready when you are
Create an account, connect a VPS, and ship your first application.