What changed in Monitorable

An ntfy server entered as http://… is now reached over http. Before, such a channel could not deliver. A user:password@ and a path in the server address are kept, so a self-hosted ntfy with access control or under a sub-path works too.

See Notification channels.

A port incident that closed as Retired after 24 hours no longer opens again when the agent goes quiet and comes back soon after. Before, a short agent outage in that window sent a new alert about a port that had already been gone for a day.

See Listening ports.

When you test a webhook or ntfy channel whose destination resolves to a private or reserved address, or answers with a redirect, the error now says which one and what to change: use a public URL, or the final URL after the redirect. Other failures still show “test notification failed”.

See Notification channels.

A webhook or ntfy destination that answers with any redirect (3xx) now counts as a failed delivery. Before, a 300 without a target and a 304 counted as delivered. Point the channel at the final URL.

The live demo opens the dashboard on a real fleet: servers running PostgreSQL, nginx and Gitea, with their own alert template and incidents. It is read-only and needs no account; every control that would change something is disabled, and server addresses are hidden.

The systemd, SMART, Docker, host-metrics and no-data pages each end with a step-by-step recipe for the alert that matches them (service fails, disk about to fail, container stops, disk filling up, agent goes quiet), and the rules page explains how to set up one alert.

See Alert rules.

Each server page lists the TCP ports its services listen on, with the number of established connections per port and a history, and a “Port not listening” rule fires when a port you rely on closes. Needs agent v1.5.0 or later.

See Listening ports.

The outdated marker on the servers list compared versions loosely and could flag an agent that was current. It now compares strict semantic versions.

If Monitorable itself stops receiving data (a deploy, an outage on our side), the gap is shown on the incident timeline as ours, and it does not fire or extend an Agent offline incident for your servers.

See Alerting overview.

Paid plans are open, with card payment through Creem. They add more servers and longer metric history; Pro adds Slack, ntfy and webhook channels, Scale and Enterprise add PagerDuty too. The free Lite plan stays.

See plans and prices.

An alert template can send each rule to its own channels, and can escalate an unacknowledged incident to a second set of channels after a delay.

See Notification channels.

Threshold rules now resolve only once the value is clearly back under the threshold (a dead band at 90 % of it), resolve notifications are held through the 15-minute reopen window so a flapping metric sends one message instead of several, and reminders pause while a server’s agent is offline.

See Alert rules.

You can mute a server for a chosen duration and note why. Incidents on a muted server are held rather than resolved, the server leaves the attention tiers and the alerts badge, and alerting resumes when the mute ends.

See Alerting overview.

The server page’s Permission limitations card now names only privileges you can actually grant, and the blanket “temperature monitoring needs root” notice is gone. Needs agent v1.4.5 or later.

See Requirements and permissions.

Monitorable is open to everyone. Install one agent per Linux server with a single command and get CPU, memory, disk, network, Docker containers, systemd services and SMART disk health in one dashboard, with alert rules that notify you by email, Telegram or Discord on the free Lite plan. The agent is a custom build of the OpenTelemetry Collector (open source, v1.4.4 at launch); it updates itself without a key and rolls back if an update fails, announces the server’s public IPv4 and IPv6 addresses, and never sends private IPs or MAC addresses.

See Getting started, What the agent collects and plans.