Skip to content
S.A.M fleet

Twenty machines, one operating surface.

A public evidence layer for the hardware behind tinyblue: doctrine from AGENT1, mesh state from Tailscale, and cluster facts from the self-hosted k3s control plane.

What we publish

The claim, without the wiring diagram.

We run a real fleet — workstations, laptops, servers, a small k3s cluster. We don't put the host map, mesh IPs, or role specifics on the public web. The counters above are the only public surface; everything else stays inside the operator boundary.

Doctrine

Machine identity, network segment, and operator notes come from the AGENT1 fleet registry — they are not exposed via the public API.

Mesh state

Online/offline counts are derived from `tailscale status --json` on the master box, aggregated before publication. Mesh IPs and relay codes never leave the operator network.

Cluster proof

Pod and node counts are the live local k3s control-plane facts. The cluster surface is publishable; the topology behind it is not.