Hermes

What Hermes Is

Hermes is the AI agent I have running permanently on the homelab. It's connected to Discord, so I can fire off questions, trigger summaries, or hand off small tasks from my phone, laptop, or wherever — without SSH-ing into anything or opening a dashboard.

The name fits. It's the messenger layer between me and everything else running on the stack.

Discord Integration

Hermes lives in a private Discord server I use as a homelab control panel. There are channels for monitoring, alerts, n8n notifications — and one dedicated channel for talking directly to Hermes.

When I send a message there, it picks it up, processes it, and replies in thread. No cold start, no waiting for an external API. It runs on the local network and responds to Ollama or Claude depending on what I've configured for the task.

Setting Up the Discord Bot

1. Go to the Discord Developer Portal and create a new application
2. Under Bot, enable Message Content Intent and Server Members Intent
3. Copy the bot token — this goes in your Hermes config
4. Invite the bot with the bot OAuth2 scope and Send Messages + Read Message History permissions

Wiring It Up

DISCORD_TOKEN=your_bot_token
DISCORD_CHANNEL_ID=the_channel_id_hermes_listens_on

Hermes watches the configured channel and replies in thread to keep things clean. Any message it doesn't understand it'll say so, which is fine — the point isn't a perfect chatbot, it's a useful one.

What I Use It For

• Service status — "is Grafana up?" without opening another tab
• Summaries — ask for a summary of recent n8n workflow runs or monitoring alerts
• Quick lookups — anything I'd normally have to SSH or dig through logs for
• Reminders — lightweight task notes I can drop into the channel and pick up later
• Homelab questions — it has context about what I run, so answers are actually useful

Why Having It in Discord Matters

Everything else I track is already in Discord — monitoring alerts, n8n workflow notifications, Uptime Kuma alerts. Having Hermes there too means I don't need another app or interface. One Discord server covers alerts, status, and the ability to ask questions. That's a much tighter loop than tabbing between dashboards.

The other reason is mobile. I can ask Hermes something while I'm out and get a useful answer without VPN-ing in or waiting to get back to a computer.

Notes

This is a living setup. The Discord connection is stable; the LLM backend and tools wired to Hermes shift as I add things to the stack. The core idea stays the same: one always-on agent, accessible from anywhere, with enough context about the homelab to actually be useful.