ActivePublished: February 10, 2026· Updated: July 1, 2026

Atlas Monitoring

A lightweight, self-hostable observability dashboard for small infrastructure fleets.

Live siteSource code

Tech stack

AstroTypeScriptCloudflareSQLite

Overview

Atlas Monitoring collects metrics and uptime data from a handful of servers and renders them in a fast, static-first dashboard. It is built for people who want a clear picture of their infrastructure without running a heavyweight monitoring stack.

Why I built it

Most monitoring tools are either too simple (a single ping check) or far too complex (a full metrics platform to operate). Atlas sits in the middle:

  • Static-first — the dashboard is prerendered and hydrates only the live widgets.
  • Self-hostable — a single container, no external dependencies.
  • Readable — sensible defaults, no dashboard-building required.

How it works

  1. Lightweight agents push metrics to a small collector.
  2. The collector stores rolling windows in SQLite.
  3. The dashboard reads snapshots at build time and streams live updates over a tiny WebSocket.

The goal was “boring on purpose” — something I can leave running for a year and not think about.

Status

Actively maintained and running in production on my own fleet.

Knowledge base

Guides and notes related to this entry.

3 articles

  • Knowledge base

    Getting started with Atlas

    Install the agent

    Drop a single binary on each host and point it at your collector:

    curl -fsSL https://example.com/atlas/install | sh
atlas agent start --collector https://atlas.example.com

    Within a minute you should see CPU, memory, and a basic uptime check on the dashboard.

    First dashboard

    1. Open the Atlas UI.
    2. Confirm the host appears under Fleet.
    3. Pin the default overview widgets — they are intentionally boring.

    This stub is placeholder copy so the modal layout can be reviewed.

  • Knowledge base

    Alerting playbook

    When Atlas pages you

    Atlas only alerts when a check fails for the full duration of its window. Short blips are rolled up, not escalated.

    Tuning noisy checks

    • Raise the failure window from 2m to 5m for flaky endpoints.
    • Silence overnight maintenance with a temporary mute.
    • Prefer host groups over per-host overrides when possible.

    Escalation

    Stub note: wire this to your pager later. For now this article exists to demonstrate the scrollable modal.

  • Knowledge base

    Self-hosting on a VPS

    Minimal Compose layout

    services:
  collector:
    image: example/atlas-collector:latest
    ports: ["8080:8080"]
  dashboard:
    image: example/atlas-dashboard:latest
    depends_on: [collector]

    Mount a volume for SQLite so restarts keep history. TLS terminates at your reverse proxy.

    Ops checklist

    • Back up /var/lib/atlas nightly.
    • Keep agents and collector on the same major version.
    • Watch disk growth on long retention windows.
Tags:infrastructuremonitoringopen-source