Hermes Agent VPS Tutorial

Status: planning Port: N/A Path: /opt/data/projects/hermes-vps-tutorial.md

What It Is

A practical tutorial for Medicare colleagues who currently use web-based AI (ChatGPT, Gemini) to level up to a self-hosted AI agent. Based on Henry's real experience.

Henry's POV

I'm not a programmer. I'm a regular person who figured this out. If I can do it, you can too. This tutorial is written from a non-tech perspective — no jargon, no assumptions about coding knowledge.

Target Audience

Medicare colleagues who only use ChatGPT/Gemini web interface
No programming background required
Basic computer skills only

Core Philosophy

"From ChatGPT to Your Own AI Agent"

Start from what they already know, then show them: 1. What they're missing (limitations of web AI) 2. What's possible (self-hosted agent) 3. How to get there (step by step)

Chapter Structure

Part 1: From ChatGPT to AI Agent

Chapter 1: From ChatGPT to AI Agent

What you know: ChatGPT, Gemini web interface
What you're doing: Chatting with AI, getting answers
The limitation:
AI forgets everything when you close the tab
AI can't do things for you (only suggests)
AI can't connect to your tools
AI can't work while you sleep
What's an AI Agent?
Analogy: ChatGPT = intern who forgets everything. AI Agent = assistant who remembers and acts.
Key difference:
- ChatGPT: You ask → It answers
- AI Agent: You ask → It thinks → It acts → It remembers
Real examples:
- "Summarize my inbox every morning" (ChatGPT can't, Agent can)
- "Monitor deals and notify me" (ChatGPT can't, Agent can)
- "Schedule meetings from email" (ChatGPT can't, Agent can)
Why Self-Host?
Privacy: Your data stays on your server
Control: You decide what it can do
Cost: Pay for API, not subscription
Future-proof: Skills you learn apply to any AI agent

Part 2: The Building Blocks

Chapter 2: LLM - The Brain

What is LLM: The AI model that understands and generates text
You already use it: GPT-4 (ChatGPT), Gemini
Options:
Cloud APIs (OpenAI, Anthropic, Google)
Local models (for privacy)
Why it matters: You can choose different "brains" for different tasks

Chapter 3: MCP - The Hands

Analogy: LLM is the brain, MCP is how it connects to hands (tools)
What it does: Standard way for AI to use external tools
Examples:
Google Calendar (check schedule)
Email (send, read)
Files (read, write)
Web (search, scrape)
Why it matters: One standard, works everywhere

Chapter 4: Skills - The Memory

What are Skills: Reusable knowledge the agent learns
Analogy: Like teaching an employee once, they remember forever
Examples:
"How to scrape deal websites"
"How to format company reports"
"How to handle email responses"
Why it matters: Agent gets better over time

Part 3: Your First Agent (Zeabur)

Chapter 5: Why Zeabur?

What is Zeabur: Managed VPS platform (like Shopify for servers)
Why I chose it:
No coding needed
One-click deploy
Built-in monitoring
Good support
Cost: ~$5-10/month
Comparison: vs DigitalOcean, Linode, self-hosted

Chapter 6: Installation (15 minutes)

What you need:
Zeabur account (free tier available)
Telegram account (for messaging)
API key (for AI model)
Basic concepts (explained simply):
VPS = your own computer in the cloud
Gateway = the messenger that connects AI to Telegram
API = how AI talks to other services
Step by step:
Create Zeabur account
Deploy Hermes (one click)
Connect Telegram
Test it works

Chapter 7: Your First Conversation

Open Telegram
Send a message to your agent
See it respond
Congratulations: You now have your own AI assistant!

Part 4: Making It Useful

Chapter 8: Teaching Your Agent

Using Skills:
How to load existing skills
How to create new skills
How to share skills with team
Examples:
"Monitor competitor prices"
"Summarize daily news"
"Format meeting notes"

Chapter 9: Scheduling Tasks

Cron Jobs: Tell agent to do things at specific times
Examples:
"Every morning at 9am, summarize my emails"
"Every Friday, send weekly report"
"Monitor website every hour, alert me on changes"

Chapter 10: Connecting Your Tools

Email: Read, send, organize
Calendar: Check schedule, create events
Files: Read documents, create reports
Web: Search, scrape, monitor

Part 5: Why This is Future-Proof

Chapter 11: The Skills You're Learning

AI Agent concept: Applies to any agent (Hermes, OpenClaw, Claude Code)
MCP standard: Used by all major AI companies
Self-hosting: Skill that becomes more valuable as AI grows
Automation thinking: How to decompose tasks for AI

Chapter 12: Your Investment

Time invested: Learning these concepts
Value gained:
Productivity boost (hours saved per week)
Career advantage (AI skills are rare)
Future flexibility (can switch agents easily)

Chapter 13: What's Next

Advanced features:
Multi-agent systems
Custom tools
Integration with company systems
Community:
Hermes Discord
Open source contributions
Sharing skills

Research Topics (To Explore)

Channel Connections

[ ] WeChat / 微信
[ ] LINE
[ ] Signal
[ ] Slack
[ ] Microsoft Teams

Mainland China Environment

[ ] VPS 選擇（國內 vs 海外）
[ ] 網絡限制（GFW）
[ ] 模型 API 連接（國內 provider）
[ ] 合規問題

Work Use Cases (Medicare)

[ ] Email 整合（IMAP/SMTP）
[ ] Calendar 整合（Google Calendar, Outlook）
[ ] 文件處理（Documents, Sheets）
[ ] 團隊協作

Online Resources (Courses)

Udemy (付費)

Hermes Agent: Build a Self-Improving AI Agent
https://www.udemy.com/course/hermes-agent-build-a-self-improving-ai-agent/
Hermes Agent Masterclass: Build Autonomous AI Workflows
https://www.udemy.com/course/hermes-agent-masterclass-build-autonomous-ai-workflows/
OpenClaw: Run Powerful & Autonomous AI Agents Securely
https://www.udemy.com/course/openclaw-run-powerful-autonomous-ai-agents-securely/
OpenClaw Tutorial: Build Powerful AI Agents (Free Course)
https://www.udemy.com/course/openclaw-tutorial-build-powerful-ai-agents-free-course/

YouTube (免費)

"The Ultimate Beginner's Guide to Hermes Agent"
https://www.youtube.com/watch?v=8yE6G1Lup1s
"Hermes Agent Tutorial: Beginner to Pro In 13 Minutes"
https://www.youtube.com/watch?v=-NF6TYwRJ2U
"Hermes Agent - Full Course & Setup Guide - For COMPLETE Beginners"
https://www.youtube.com/watch?v=mTYxpIRK7xA
"Deploy Your Own AI Agent in 45 Minutes | Beginner OpenClaw Tutorial"
https://www.youtube.com/watch?v=sO6NSSOWDO0

Guides

Lenny's Newsletter: "OpenClaw - The complete guide to building, training, and living with your personal AI agent"
https://www.lennysnewsletter.com/p/openclaw-the-complete-guide-to-building

Notes

Henry's setup: Zeabur VPS, Telegram + WhatsApp, n8n for Google services
Consider both Zeabur (managed) and self-hosted options
Include cost breakdown if possible
Write in Traditional Chinese (繁體中文) for HK audience
Gap: No course covers "real business use cases" — this tutorial can fill that gap