Infrastructure as Readme for Your Agent Fleet

What it does Hermes Agent Control Room is a sidecar template for governing Hermes agents on a VPS. It provides a folder structure, runbook templates, and bundled skills for Claude Code or Codex to bootstrap servers, register agents, and manage backups. The repo itself is not an agent; it is the system map and operating manual that agents (and you) consult.

The interesting bit The unusual discipline here is setting up the bureaucracy before the workers. The template enforces a four-level maturity model—one agent, then direct specialists, then an orchestrator, then automation—so you only add complexity after the manual workflow actually works. It also keeps raw secrets out of the control plane by design, recording only metadata like scope and rotation dates.

Key highlights

• Sidecar control plane: docs, registry, and runbooks live in /root/agent-control-room while runtime state stays in /srv.
• Bundled skills for agent-assisted setup, including a Hetzner VPS bootstrap and an agent-task-router for orchestrator handoffs.
• Task bus pattern for routing work between an optional orchestrator and specialist agents.
• Security-by-default templates: inventory, env maps, and backup plans without committing API keys.
• Four-level architecture that delays automation until the manual system is debugged.

Caveats

• This is a template and skill library, not a running service; your agent still has to execute the setup steps interactively.
• The Hermes Agent installation is marked “best-effort,” and several bundled skills assume a Claude Code or Codex workflow.
• The automated VPS bootstrap is tightly coupled to Hetzner for the initial create-vps skill.

Verdict Grab this if you are running (or planning) multiple Hermes agents on a VPS and want structured governance before the chaos sets in. Skip it if you are looking for a hosted multi-agent orchestrator or a one-click runtime.