I got tired of SSH-ing into six machines, so I built an open-source homelab dashboard

I run a small homelab with Linux boxes, Windows machines, ARM devices and a few GPU systems The annoying part was never any one machine. It was managing all of them Something feels wrong, so you SSH into one box, check CPU, move to another, check disk, try to remember an IP, realize another machine is offline… and 20 minutes later you still don’t know what actually happened So I started building BloxOS. It gives me one dashboard for all my machines: • CPU, RAM and disk usage • GPU monitoring • power readings • alerts • uptime / online status • Linux web terminals • services • basic AI-session monitoring for tools like Claude Code, Codex and Kimi The agents connect outbound to the hub, so I don’t need to open inbound ports on every machine The whole thing is self-hosted. The hub is Go + SQLite, the frontend is Next.js, and there’s no Redis, Kubernetes or external database One rule ended up shaping a lot of the project: If BloxOS doesn’t know something, it shows N/A instead of guessing. That sounds obvious, but monitoring software gets surprisingly dangerous when missing data silently becomes 0, or estimated numbers look like measured ones I also wrote up some of the bugs I hit while building it, including: • one screen showing 1 alert and 0 alerts at the same time • a theme bug that produced literal black text on a black background • an Orange Pi RK3588 being reported as 8 Cortex-A55 cores instead of 4x A76 + 4x A55 • two identical GPUs breaking a React list because I used the GPU model as the key All of them passed CI in one way or another. BloxOS is still very much a personal project. I currently run it across 12 machines, and I already know some areas that will need work if people start running much larger fleets But it’s open source, Apache 2.0, and I’d genuinely like feedback from other homelab people repo: https://github.com/bokiko/bloxos