What is homelab-monitor?

HomeLab Monitor is a single-container, self-hosted dashboard that puts your entire home lab and AI rig on one page. It shows live GPU utilisation, VRAM, power draw, temperature and identifies which container or process is using each GPU — supporting NVIDIA, AMD and Intel GPUs on Linux and Windows. It tracks Docker container health with separate RAM and VRAM columns, systemd services, interactive disk treemaps, network interfaces and listening sockets, plus a read-only security posture check covering firewall, SSH, SELinux/AppArmor and fail2ban. Power consumption is measured per component and per process, container or model, with day/night tariffs and a weekly busy-hours heatmap. A Benchmark Lab sweeps local Ollama models across context sizes to find the largest context fitting in VRAM. Multi-machine monitoring works agentless over SSH — just add the hub's key to each remote (Linux, Raspberry Pi, or Windows). A built-in read-only MCP server on port 9810 lets AI agents explore the homelab through 19 tools. Alerts go to Discord, ntfy.sh, Telegram, Email, Slack or generic webhooks. Runs without a separate metrics stack or cloud dependency, no login required — meant for a trusted LAN/VPN. History persists in SQLite (./data/gpu.db).

Should you self-host homelab-monitor?

homelab-monitor scores 52/100 on selfhost.directory (how scoring works) and ranks #19 of 51 in Monitoring & observability. It has 227 GitHub stars.

It shipped 23 stable releases in the past year. The latest, v0.37.0, came out on 23 September 2026. New versions land roughly every 2 days.

Setup is moderate: expect some configuration beyond a single command. An official Docker image is available. It is MIT-licensed and written mostly in Python.

Its closest alternative on selfhost.directory is GoAccess (75/100, 21 091 stars). It scores higher overall, so compare the two before you pick. See homelab-monitor vs GoAccess side by side.

Features

AI model awareness
Alerts
Benchmark Lab
Container MonitoringControls can be disabled with ENABLE_CONTROLS=0
Disk treemaps
Experiment tracking
GPU Monitoring
MCP server (read-only)Can be disabled with ENABLE_MCP=0
Multi-host over SSHContainers tab remains local-only for now
Network Monitoring
Power and cost tracking
Prometheus metrics export
Security posture checkRead-only, no remediation
Systemd service monitoringControls can be disabled with ENABLE_CONTROLS=0; remote Windows via PowerShell
Uptime Monitoring

How to configure homelab-monitor

ALLOW_SELF_UPDATE1Set 0 to disable in-app updating
Set 0 to disable in-app updating
values: 0,1
CHECK_OS_UPDATEStrueSet false to stop reporting pending OS package updates
Set false to stop reporting pending OS package updates
values: true,false
CHECK_UPDATEStrueSet false to disable the daily GitHub-releases check (no outbound calls)
Set false to disable the daily GitHub-releases check (no outbound calls)
values: true,false
DB_PATH/data/gpu.dbWhere history is stored inside the container
Where history is stored inside the container
DOCKER_SOCK/var/run/docker.sockDocker socket to read containers from
Docker socket to read containers from
ENABLE_CONTROLS1Set 0 to remove the start/stop/restart buttons from the Containers and Services tabs
Set 0 to remove the start/stop/restart buttons from the Containers and Services tabs
values: 0,1
ENABLE_MCP1Set 0 to run the dashboard without the MCP server
Set 0 to run the dashboard without the MCP server
values: 0,1
FAST_INTERVAL2Seconds between live-value refreshes on screen. Reads only cheap counters and stores nothing. 0 turns the push stream off and the dashboard falls back to polling.
Seconds between live-value refreshes on screen. Reads only cheap counters and stores nothing. 0 turns the push stream off and the dashboard falls back to polling.
HOST_ROOT/rootfsMount point of the read-only host root
Mount point of the read-only host root
MCP_PORT9810Port for the built-in read-only MCP server
Port for the built-in read-only MCP server
PORT9800Dashboard port
Dashboard port
PRESSURE_FREE_MB2048Free VRAM below this counts as pressure
Free VRAM below this counts as pressure
PUBLIC_STATUSSet to enable the public status page (also a Settings toggle, which needs no restart)
Set to enable the public status page (also a Settings toggle, which needs no restart)
RETENTION_DAYS180How long history is kept
How long history is kept
SAMPLE_INTERVAL10Seconds between stored samples. This is the storage cadence — every energy and cost figure is integrated against it.
Seconds between stored samples. This is the storage cadence — every energy and cost figure is integrated against it.
WATCH_CONTAINERSExtra containers to scan for OOM (comma-separated)
Extra containers to scan for OOM (comma-separated)
WATCH_SERVICESsystemd units to always show, even vendor ones (comma-separated)
systemd units to always show, even vendor ones (comma-separated)