Agent dashboards optimize the session
Most agent dashboards focus on starting sessions, reading output, sending prompts and receiving notifications. That is valuable, especially when several agents work in parallel. Their blind spot is usually the host around the session: disk pressure, failed services, Kubernetes state, traffic and operational evidence.
Infrastructure consoles optimize the runtime
Kubernetes and container dashboards are excellent at workloads, resources and logs. They generally do not understand that a tmux pane contains a coding agent waiting for a human, or that several agents are changing different parts of the same platform.
Server Harbor joins the two views
Server Harbor keeps session control close to host and cluster visibility. The current implementation includes health, capacity, services, Kubernetes workloads, application traffic, curated logs, safe transfers, documentation and a Claude Code/Codex-aware tmux cockpit.
It also makes a different security choice from a typical remote admin panel: the service binds to loopback and is designed for SSH-tunnel access. Arbitrary command execution and destructive infrastructure buttons are intentionally absent.
Who it is for
The clearest fit is a developer or team operating coding agents alongside real services on infrastructure they control. It is particularly relevant where agent access, audit evidence and operational confidentiality matter more than public-anywhere convenience.