Linux server monitoring that stays on your servers

Servers get the agent package, moxivue-agent: the agent, the read-only helper and the moxivue command line, without any window or desktop libraries. It needs Ubuntu 22.04 or newer (tested on 22.04, 24.04 and 26.04), for Intel/AMD and ARM64.

sudo apt install moxivue-agent
sudo moxivue service enable
sudo moxivue machine join <rendezvous>
sudo moxivue verify

The rendezvous comes from an owner's or admin's console (moxivue machine add). The server shows an 8-digit number: type it on that console, which then shows three words to type on the server. For many servers at once, a script joins each with an install token and the account's full fingerprint: sudo moxivue machine join --token --fingerprint <fingerprint>.

The agent runs as the system user moxivue, which MoxiVue creates, so it starts at boot with nobody signed in, and the moxivue commands on the server run with sudo. It starts again by itself after a crash or a kill, and a server that sends no signed status for 5 minutes shows as silent to the account's owners and admins. A server never has a role in its account: only the consoles of the people who manage it do. moxivue verify checks every part and exits with status 1 if something does not work. The guide has the full steps.

• A reading every 30 seconds into the history by default (5 seconds to 5 minutes), and every 2 seconds while someone has its live view open: processor per core, memory, disk activity and free space, network, temperature and pressure (how long apps were stalled waiting for memory, disk or processor).
• Every process, with its processor, memory, disk and network use.
• Drive health from SMART, checked hourly; systemd services; Docker and Podman containers and virtual machines.
• History kept for 30 days by default (adjustable), with per-app totals in 30-minute blocks, and export as CSV.

moxivue status shows the machine at a glance, with --more or --all for detail and --json for scripts.

The agent checks every 30 seconds for problems: disks full, memory, temperature, drive health, failed services, kernel log errors (out-of-memory kills, disk I/O errors, file systems gone read-only), links down, gateway, internet and DNS problems, certificate expiry and more. Each fault has a severity, first and last seen times and a count, and stays on the list until someone acknowledges, mutes or clears it. On a server, moxivue faults lists them.

Alerts cover processor, memory, disk space, temperature, pressure, drive health and activity far from the machine's normal for that hour of the week (after 7 days of history). They go to ntfy, Telegram, email, SMS or WhatsApp, with a repeat interval so a lasting problem doesn't flood you. The feature tour lists every alert and its default.

Each server connects out to the MoxiVue Dispatcher, so no port is opened. Your other machines show it under Machines, online or last seen, with its full dashboard one click away.

The server seals its readings, status and audit records with keys only the consoles of its site or group hold, and sends a signed status every 2 minutes and records every 5 minutes; the Dispatcher stores them sealed. What the Dispatcher can see, such as your sign-in email, which devices belong to your account and when they are online, is listed on the security page.

A shell on the server, from your console, in a tab or with moxivue terminal <server>. Operators, admins and owners whose site or group covers the server may open one from an unlocked console. Terminals are off until an owner or admin switches them on for the account, a site, a group or this server, and a second console confirms; the setting can name the consoles that may connect. A server has nobody to answer an ask-each-time prompt, so it goes by that list. Root on the server can refuse sessions there altogether (sudo moxivue machine kinds off session.start).

• Never as root: the shell runs in the account moxivue-session, and every sudo there needs a yes on the console that opened the session.
• Encrypted between your console and the server with keys made for that connection.
• The server checks your console's role against its own copy of the account's log when the session opens and every 30 seconds after.
• Sessions end after 15 minutes without typing or 8 hours in all.
• The server signs its own record of who connected, when and for how long, and records the session itself, sealed for the owners, admins and auditors of its place; a session that cannot be recorded does not start.

If you prefer your own ssh, an owner switches SSH from the account on for the server (moxivue ssh-host on --machine <server>, a second console confirms) and grants your console's key for up to 30 days (moxivue ssh grant); then moxivue ssh <server> runs your ssh over your console's connection. The key goes into your own account on the server, so this needs per-person accounts switched on by an owner. Sessions with that key are recorded like terminal sessions. How it is locked down is on the security page.

sudo apt update and sudo apt upgrade, or moxivue update, which shows the new version and asks. An owner or admin whose place covers the server can also ask from the console (moxivue update <server>): its root executor checks the request against its own copy of the log and upgrades only MoxiVue's own package from the package source already on the server. apt checks every package against the MoxiVue key on the server.

MoxiVue is native code written in Rust. On our cloud server with one processor core, the agent used 24 MB of memory and the helper 6 MB, and both together 2.2% of the processor, averaged over 45 minutes with the default settings. The amd64 agent package is an 11.4 MB download (39.5 MB installed). Traffic with the Dispatcher was about 1.4 MB a day per machine. Your figures will differ; the measurements say how to check them.