Zuvio Atlas

Server monitoring that installs in one command

One command installs a 10 MB agent that reports CPU, memory, disk, network and Docker containers every 60 seconds — and finds the services already running without being told about them.

What you get, 60 seconds after you paste the command

A single static binary starts reporting on a 60-second cycle. No configuration file, no YAML, no dashboard to assemble first — the server appears in your fleet view already populated.

dashboard / servers

Agents online

4 / 4

Alerts firing

0

Avg CPU

33%

Last telemetry

12s ago

web-01
nginxdockerd
db-primary
postgres
worker-02
redisdockerd
web-02
nginx

cpu / mem · 60s batch · services rescanned every 5 min

Every host, its live CPU and memory, and the services the agent found on its own.

The metrics, specifically

Sampled every 5 seconds and sent as a batch every 60, so a spike between batches is still in the data rather than averaged away at the source.

  • CPU — total and per-core utilisation, plus 1/5/15-minute load averages.
  • Memory — used, available, cached, and swap.
  • Disk — usage and free space per mounted filesystem, plus read/write throughput and IOPS.
  • Network — bytes and packets in and out per interface, plus established TCP connection counts.
  • Processes— the top 15 by CPU each cycle, so “what was actually running at 03:00” is answerable after the fact.
  • Docker — per-container CPU, memory, network and block I/O, with restart counts.

It finds what is running without being told

Most monitoring makes you declare your services before it will watch them, which means the one nobody remembered to declare is the one that fails silently. The agent enumerates what is actually running on the host instead — nginx, PostgreSQL, MySQL, Redis, Docker, and anything else holding a systemd unit — and rescans every five minutes.

A service installed a month after the agent still turns up. So does a service that stopped: the discovery list is replaced on each report, not merged, so a process that disappears disappears from the dashboard too rather than lingering as a permanently green entry that nothing is checking.

The same pass finds scheduled jobs in the host’s user crontabs, which is how cron monitoring gets set up without anyone having to remember what is in there.

Safe to run on production — measured, not asserted

An agent that costs you an incident is worse than no agent. The numbers below are measured on real hosts, not modelled:

  • Under 10 MB of RAM and under 0.5% of one CPU, steady state. It is a static binary with no runtime, no interpreter and no package dependencies.
  • No root. It runs as a dedicated unprivileged user under systemd sandboxing — ProtectSystem=strict, NoNewPrivileges, read-only elsewhere.
  • No inbound ports. Outbound HTTPS on 443 only, so there is no firewall rule to add and it works behind NAT.
  • It reads, it does not act. No command execution, no state modification. The single exception is optional, off by default, and is one read-only systemctl show call used to observe job outcomes.
  • Signed updates. Every binary is Ed25519-signed and SHA-256 verified before it is swapped in; a failed check aborts rather than degrading.

A network drop is not a data gap

When the control plane is unreachable the agent does not discard telemetry. It buffers to a local ring, persists that buffer to disk so a reboot mid-outage does not lose it, and drains on reconnect at a rate that respects the ingestion rate limit. Backoff runs from 2 up to 16 minutes while the API stays down.

Why this matters

The window you most want data from is the one where things were going wrong — and that is exactly when a network path is most likely to be flapping. Monitoring that drops telemetry during an incident is missing the incident.

Install

Linux (Ubuntu, Debian, Fedora, CentOS, RHEL, Rocky, Alpine, Amazon Linux) on x86_64 and aarch64, plus Windows Server on x86_64. The command below is read-only and installs nothing — it reports what Atlas would see on the host.

check your server — read-only

$curl -fsSL https://atlas.zuviosystems.com/install.sh | sh

Explore more

14-day free Pro trial — no credit card required

Everything monitored. One bill.

Full Pro access for 14 daysNo credit card requiredDrops to the free plan if you don't upgrade