Dumb-simple, self-hosted monitoring for homelabs and small teams. One container, one IP address to type in, useful graphs and alerts in under a minute.
DumbMonit$ mkdir dumbmonit && cd dumbmonit
$ curl -fsSLO https://raw.githubusercontent.com/noekan/dumbmonit/main/docker-compose.yml
$ docker compose up -dThen open http://localhost:8080 and create the first account. Image for amd64 (arm64 later): ghcr.io/noekan/dumbmonit.
This is an alpha. DumbMonit runs every day on its author's own network, but the HTTP API is not frozen, the database schema still moves and there is no upgrade guarantee across schema changes yet. Several integrations have only met test instances, not years-old production hardware; they are marked below. Try it, and tell us what breaks.
Pick a type, read the notice that says what to prepare on the device, fill in only the fields that type needs. SNMP profiles are picked from the device's sysObjectID; defaults do the rest.
Every kind has its own page in the device documentation, with exactly what it reads and what to prepare. A demo device kind produces data without any hardware.
Useful rules are on right after install. The work goes into sending fewer, better messages.
Declare a switch as the parent of what sits behind it. When it goes down you get one notification, not thirty; the others read “Suppressed by parent”.
A NAS whose array degrades sends one message per cycle, not one per disk. Reminders, escalation and acknowledgement (“I know, stop reminding me”) are built in.
Anomaly detection with no threshold to tune: 168 buckets, one per hour of the week. It stays silent for its first 14 days and only shows what it would have fired.
“This filesystem will be full within four days” is listed as a forecast, apart from what is broken right now.
One-off or weekly windows, hysteresis, flap hold, quiet hours, batching and an hourly cap, so a bad night does not become two hundred pushes.
Each one has a Test button and a setup page in the documentation.
A small binary that pushes CPU, memory, disks, network, services and uptime to the server, plus temperatures, fans, SMART and ZFS pools where the machine exposes them. It calls the server, never the other way round: no port to open on the machine.
# the UI fills in the token for you
curl -sSL http://server:8080/install.sh | sh -s -- \
--token=dmon_xxx --url=http://server:8080A PowerShell script is served at /install.ps1 for Windows. The agent is also an image, ghcr.io/noekan/dumbmonit-agent.
In relay mode an agent runs the server's probes (SNMP, Proxmox, HTTP…) from its own network, so a second house or a small office shows up in the same DumbMonit without a VPN. Monitor a remote site
Public pages at /s/<name>: groups of services with 90 days of daily history, 30- and 90-day uptime, incident and maintenance announcements, email subscribers with double opt-in, RSS, badges for a README and a compact embed for an intranet.
Only the labels you choose leave the server. Device names, addresses and types never appear on the page, in a badge or in an email.
The bulletin alone, full screen, for a room monitor. It keeps the screen awake.
Admin and viewer roles, optional TOTP second factor, an audit log, and single sign-on over OpenID Connect (Authentik, Authelia, Keycloak, Pocket ID…).
/metrics, /federate and a Prometheus-compatible endpoint let an existing Grafana read DumbMonit's data. A REST API with scoped tokens, and a built-in MCP server for AI assistants.
Read the state, acknowledge an alert, start a maintenance window. Installable on the home screen, deliberately without an offline mode: it never shows yesterday's state.
One Rust binary with the web interface built in, the VictoriaMetrics time-series database started by that binary inside the same image, and SQLite for configuration. No database container, no message queue.
NET_RAW.DUMBMONIT_VM_URL at it and the embedded one is not started.DumbMonit is for people who watch ten machines and three switches. The established tools are excellent at what they do; here is where each one fits better.
Mature, proven on thousands of hosts, with a template for nearly everything, distributed proxies and long-term support releases. It also asks for a database server, a web front end and time to learn items, triggers and templates.
Choose Zabbix for large fleets or when you need that depth today. Choose DumbMonit when you want the one-second answer and can live with fewer integrations and a young project.
The standard for metrics: a huge exporter ecosystem, PromQL and dashboards for anything, including your own applications. You assemble and maintain the stack yourself: Prometheus, Alertmanager, exporters, dashboards and rules.
DumbMonit ships that assembly already done, and exposes /metrics, /federate and a Prometheus-compatible endpoint, so the two can live side by side.
A polished, very popular uptime monitor with more monitor types, more notification services and a larger community. It checks services from the outside; it does not read a switch's ports, a hypervisor's guests or a NAS's disks.
DumbMonit includes Kuma-style checks and push-compatible heartbeats, next to device metrics. If uptime checks are all you need, Uptime Kuma is the more mature choice.
Apache 2.0, dependencies included: CI fails on any dependency that is not under an OSI-approved licence. No feature is held back for a paid edition, because there is none.
This site has no analytics, no cookies and no third-party requests.
A device that reads wrong, an alert that woke you for nothing, a page that confused you: each report makes the next alpha better. Device support requests are welcome too.