About Glassmkr
Monitoring for bare metal, built and run by one operator.
Who we are
I have spent a decade operating bare metal: racks, switches, IPMI consoles, 3 AM pages about failing drives. I know what breaks and how it breaks, which is the only reason the alert rules are opinionated rather than generic.
Glassmkr is built and operated by one person. The hosted service runs on infrastructure I administer myself.
Why we exist
General-purpose monitoring tools like Datadog and New Relic do not understand IPMI, SMART, or RAID. They are built for cloud workloads. Traditional tools like Nagios and Zabbix can monitor hardware, but they are heavy to deploy and require extensive configuration.
I wanted something opinionated: 70 alerts covering what actually breaks, with defaults rather than a configuration exercise. Setup is one command on Ubuntu or Debian and a few minutes by hand elsewhere; it needs an account key, and the agent needs root once to install before it runs as its own unprivileged user. In August 2026 I open-sourced the entire stack.
What it is
- Crucible the agent
Open-source monitoring agent for bare metal servers. 70 opinionated alert rules covering IPMI, SMART, RAID, OS, and network. MIT licensed.
- Dashboard the dashboard
Monitoring dashboard for bare metal fleets. Receives health data from Crucible, stores history, displays fleet health, sends alerts. Open source (AGPL-3.0-only) and free: hosted with a 10-node per-account cap, or self-hosted with no node limits.
- Furnace the AI assistant
Our AI assistant for the alert-remediation workflow. Reads the alert and its evidence, suggests a fix, and hedges what it does not know. A value-add in the workflow, not a separate product.
How we think
- 01
Built from operations, not theory
Every alert, threshold, and diagnostic comes from real operational experience running bare metal.
- 02
Encounter first, then build
Features come from problems I have actually hit. If I have not hit it, I do not build for it.
- 03
Transparency by default
Open-source agent and dashboard. Public documentation. No black boxes in your monitoring stack.
- 04
Craft over speed
Understanding a failure mode properly takes longer than shipping a threshold, and is the part that matters.
- 05
For operators, by an operator
Hosting companies, MSPs, sysadmins, homelabs. I run the same infrastructure you do.
- 06
Free and open, both layers
The agent and the dashboard are both open source. Run the hosted service, or run the same code on your own hardware.
The name
Glassmkr takes its name from the Bohemian glassmaking tradition: objects that are difficult to make and ordinary to use. That is the intended shape of this software too.
Contact
- General: [email protected]
- Security: [email protected]
- Privacy: [email protected]
- Code: github.com/glassmkr