Glassmkr Documentation

Open-source monitoring for bare metal. Start below, then follow the guide for whichever way you are running it.

#Start here

Two decisions, in order.

  1. Choose where the dashboard runs. Self-host the whole stack from one compose file, with no node limit and no license key, or use the hosted service and skip operating it. Self-hosting guide Hosted signup
  2. Install the agent on each machine you want to watch. Ubuntu and Debian have a one-line installer; everything else uses the single-file binary, which needs no Node. Getting started
  3. Confirm the first snapshot arrives, then route notifications and tune what fires. Notification channels Configuration

#Guides

  • Getting startedInstall the agent three ways, and which one your distribution needs.
  • Self-hostingThe compose stack, backups, upgrades, retention, and pointing agents at your own dashboard.
  • ConfigurationEvery field in crucible.yaml, with defaults and bounds.
  • Notification channelsWhere alerts go, and what each channel needs.
  • Alert rulesAll 72 rules, what each one watches, and the remediation it carries.
  • Programmatic APIAccount keys, scopes, rate limits, idempotency, and error shapes.
  • Automated onboardingEnrolling hosts from Ansible, Terraform or a script.
  • MCPQuerying your fleet from an AI tool, and what self-hosting it requires.
  • TroubleshootingWhen the agent will not start, or snapshots do not arrive.
  • FAQThe questions that come up before people install anything.
  • ChangelogWhat changed, by release.

#For machines

Every guide is served as Markdown at its own URL with a .md suffix, and every alert rule at /docs/rules/<rule_id>.md. The whole corpus in one file is at /llms-full.txt, indexed by /llms.txt. Fetch those rather than scraping HTML.