What is Hermes Agent (and why your coding agent isn't enough)

Open Claude Code right now and ask it what you were working on last week.

It doesn't know. It never did. Every session starts from nothing — no memory of your project, your preferences, or the bug you spent two hours on yesterday. That's fine for a coding session at your desk. It's the wrong tool for everything else.

Hermes Agent is built around a different premise: what if your AI assistant was always running, always reachable, and got smarter the longer you used it?

What "persistent" actually means

A session-based agent like Claude Code is a conversation. You open it, you work, you close it. The next time you open it, it's a stranger.

A persistent agent is more like a colleague who's been on your team for six months. It knows what you're building. It knows how you like things explained. It remembers that you hate boilerplate and prefer terse answers. When you ask it something, it connects the dots to things you worked on before.

Hermes achieves this through three layers:

1. Tools (70+ built-in). Terminal access, web browsing, file read/write, image generation, voice, code execution. These are available from the first session — you don't build them, they're just there.

2. Skills (procedural memory). Reusable playbooks the agent creates from experience. Every time you complete a task, Hermes can turn what worked into a skill — a step-by-step procedure it can run again next week without you explaining it. Over 600 community skills are available at agentskills.io.

3. Memory (who you are). Two things running in the background: a growing model of your preferences and habits, and a full-text searchable history of every session. Ask it "what were we debugging last Tuesday?" and it finds it.

The closed learning loop

This is the phrase Nous Research uses to describe what makes Hermes different. It works like this:

You do work → Hermes saves what worked as a skill
       ↑                                    ↓
Hermes gets better ← Curator improves skills over time

The curator is a background process that reviews skills, updates ones that have gotten stale, and creates new ones from your sessions. The longer Hermes runs, the more precisely it reflects your workflows — not some generic "helpful AI assistant" template.

Where Hermes lives

Hermes runs as a daemon — a background process that stays up even when you close your terminal. It can live on:

• Your laptop (what this course uses to start)
• A $5/month VPS like Hetzner or Hostinger (what most serious users run)
• Serverless infrastructure like Modal or Daytona (nearly free when idle)

The VPS option is the one most people end up at. When Hermes lives on a server, it's always on — even when your laptop is asleep, your phone is dead, and you're on a flight. The Discord integration in Lesson 4 is where this clicks: you send a message from your phone, and Hermes on a server picks it up, does the work, and replies.

What you'll build in this course

By the end of Lesson 6 you'll have:

• Hermes installed and talking to you in your terminal
• A SOUL.md that gives it your voice and working style
• A Discord bot that responds to your DMs from any device
• At least one community skill and one MCP server connected
• Session memory active and the Obsidian memory pattern explained

The whole thing takes about an hour. Let's start with the install.