by CryptoDmitry
Control Room-first template for managing Hermes agents from one VPS agent to specialist teams and orchestrated workflows
# Add to your Claude Code skills
git clone https://github.com/CryptoDmitry/hermes-agent-control-roomGuides for using ai agents skills like hermes-agent-control-room.
Last scanned: 5/19/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-05-19T07:45:22.356Z",
"semgrepRan": false,
"npmAuditRan": true,
"pipAuditRan": true
}A public template for setting up an Agent Control Room first, then scaling from one Hermes agent to direct specialists, orchestrated teams, and automated workflows.
The Agent Control Room is a sidecar repo/folder that documents and governs your Hermes agents. It is not an agent itself. It is the system map, operating manual, registry, runbook library, and recovery notebook for the agents you run.
It gives you a clean path from:
one agent -> direct specialists -> orchestrator -> automated agent team
Hermes Agent Control Room is a starter kit for people who want to run Hermes agents like an operating system instead of a pile of disconnected bots.
The repo gives you:
The key idea is simple: set up the Control Room first, then plug agents into it.
Create a VPS or choose an existing one.
Bootstrap the Agent Control Room.
Register one Hermes agent.
Add direct specialists when roles become clear.
Add an orchestrator when you want one front door.
Automate only after the manual system works.
The Control Room sits on the side as the control plane. You can use it directly, talk directly to any agent, or talk to an orchestrator that delegates to specialists.
Agent Control Room = side control plane
Orchestrator = optional manager/front-door agent
Specialists = focused Hermes agents with role-specific tools
Task Bus = handoff desk between orchestrator and specialists
You = owner/operator with direct access to everything
No comments yet. Be the first to share your thoughts!
flowchart LR
user["You / Operator"]
control["Agent Control Room<br/><code>/root/agent-control-room</code><br/><br/>side control plane<br/>docs / rules / registry<br/>ports / env maps<br/>runbooks / backups"]
orch["hermes-orchestrator<br/><br/>optional front door<br/>delegation / synthesis"]
bus["Agent Task Bus<br/><code>/srv/agent-bus</code><br/><br/>inbox / working<br/>outbox / archive"]
life["hermes-life<br/>personal agent"]
seo["hermes-seo<br/>SEO specialist"]
dev["hermes-dev<br/>code / site work"]
cmo["hermes-cmo<br/>marketing"]
ops["hermes-ops<br/>VPS / backups / security"]
user -->|"control path<br/>edit docs / rules"| control
user -->|"orchestrated path"| orch
user -->|"direct path"| life
user -->|"direct path"| seo
user -->|"direct path"| dev
user -->|"direct path"| cmo
control -. "defines / documents / governs" .-> orch
control -. "defines / documents / governs" .-> life
control -. "defines / documents / governs" .-> seo
control -. "defines / documents / governs" .-> dev
control -. "defines / documents / governs" .-> cmo
control -. "defines / documents / governs" .-> ops
orch -->|"routes tasks"| bus
bus --> seo
bus --> dev
bus --> cmo
bus --> ops
seo -->|"results"| bus
dev -->|"results"| bus
cmo -->|"results"| bus
ops -->|"results"| bus
bus -->|"summaries / artifacts"| orch
orch -->|"final synthesis"| user
You are never locked into one workflow.
Control path:
You -> Agent Control Room
Direct path:
You -> hermes-seo
You -> hermes-dev
You -> hermes-cmo
Orchestrated path:
You -> hermes-orchestrator -> Agent Task Bus -> Specialists -> You
Set up the Control Room and register one Hermes agent.
Best for:
You do not need an orchestrator or task bus yet.
Add multiple role-specific Hermes agents and talk to them directly.
Examples:
hermes-lifehermes-seohermes-devhermes-cmohermes-opsThe Control Room documents all of them. You choose which agent to talk to.
Add hermes-orchestrator as an optional front door. You can still talk directly to specialists, but the orchestrator can route and synthesize work.
The orchestrator follows the Control Room. It should not become a giant agent with every credential.
Add recurring workflows, audits, backup checks, task routing, and optional direct gateway/API calls.
Only add automation after the manual workflow works.
This repo includes skills that can be linked into Claude Code or adapted for Hermes.
create-vps
Create a fresh Hetzner VPS, SSH key, SSH alias, and local provisioning folder.
setup-control-room
Bootstrap an SSH-accessible VPS with Node, Claude Code, Codex, Docker,
Hermes Agent, and this Control Room template.
agent-control-room
Manage the Control Room docs and agent folders.
agent-task-router
Route tasks from an orchestrator to specialists through a task bus.
agent-registry-manager
Maintain the agent registry.
agent-backup-manager
Design and audit per-agent backups without committing secrets.
agent-security-auditor
Check ports, dashboards, SSH, Docker, secret placement, and key scope.
agent-team-cron-planner
Plan recurring multi-agent workflows after manual workflows work.
agent-control-room/
README.md
agents/
.gitkeep
docs/
architecture.md
levels.md
naming.md
security.md
task-bus.md
orchestrator.md
starter-guide.md
shared/
api-keys-sop.md
commands.md
security.md
templates/
agent/
inventory.md
docker.md
env-map.md
runbook.md
backup.md
docker/
docker-compose.agent.yml
docker-compose.orchestrator.yml
task-bus/
agents.yaml
task-template.md
result-template.md
skills/
create-vps/
setup-control-room/
agent-control-room/
agent-task-router/
agent-registry-manager/
agent-backup-manager/
agent-security-auditor/
agent-team-cron-planner/
examples/
level-1-control-room-one-agent/
level-2-direct-specialists/
level-3-orchestrator-specialists/
level-4-automated-team/
There are three ways to use this repo.
This repo is designed to be agent-readable.
If your agent can read a GitHub repo or a local clone, point it here and ask:
Read this repo and help me set up an Agent Control Room.
Start with docs/starter-guide.md and the setup-control-room skill.
If the bundled skills are available to the agent, you can be more direct:
Use setup-control-room to bootstrap my VPS.
Or, if you need a new Hetzner server first:
Use create-vps, then chain into setup-control-room.
The intended agent flow is:
create-vps
-> creates a Hetzner VPS, SSH key, and SSH alias
setup-control-room
-> installs tooling and clones this repo onto the VPS
agent-control-room
-> helps register and manage agents inside the Control Room
Important: the repo does not magically run code when opened. It gives your agent the setup instructions, templates, and skills. You still ask the agent to run the setup flow.
Use this if you already have an Ubuntu/Debian VPS you can SSH into.
SSH in:
ssh root@YOUR_SERVER
Clone the Control Room:
git clone https://github.com/shannhk/hermes-agent-control-room.git /root/agent-control-room
cd /root/agent-control-room
Read the starter guide:
cat docs/starter-guide.md
Register your first agent:
mkdir -p agents/hermes-life
cp templates/agent/*.md agents/hermes-life/
Then fill in:
agents/hermes-life/inventory.md
agents/hermes-life/docker.md
agents/hermes-life/env-map.md
agents/hermes-life/runbook.md
agents/hermes-life/backup.md
Keep raw secrets out of those files.
Use this if you have an SSH alias already configured locally.
The bundled setup-control-room skill is meant to:
/root/agent-control-room~/.claude/skillsAfter it runs, SSH into the VPS and finish interactive auth:
ssh <alias>
claude /login
codex
hermes
Then start using the Control Room:
cd /root/agent-control-room
cat README.md
ls templates/agent/
ls skills/
Do not start by building a whole agent team.
First milestone:
1. Control Room exists on the VPS.
2. One agent is documented in agents/<agent-name>/.
3. No raw secrets are in the repo.
4. You can restart/debug/recover that one agent using its runbook.
Then move to Level 2 and add direct specialists.
Keep the control plane separate from live runtime state.
/root/agent-control-room
docs, templates, runbooks, registry, architecture
no raw secrets
/srv/<agent-name>/data
live Hermes runtime
.env, memory, skills, sessions, crons, logs
Never commit raw secrets.
The control room may record:
It must not record:
MIT. See LICENSE.