Hermes Agent: The Ultimate Beginner’s Guide

One Sentence Summary

Caleb demonstrates deploying Hermes AI on a VPS, configuring OpenRouter/Telegram, memory and skills, then scheduling automated tasks across platforms seamlessly.

Main Points

• Hermes is a 24/7 self-improving AI on a server.
• It writes its own skills by observing task workflows and outcomes.
• It remembers preferences and work style over time.
• It supports scheduling tasks via natural language.
• It works across 15+ messaging platforms via a single gateway.
• A VPS is recommended for reliability, security, and always-on operation.
• Hostinger offers a one-click Hermes deployment template.
• The setup includes a detailed guide and budgeting tips.
• You can redeploy or wipe the VPS without affecting your main device.
• The video covers skills, memory, automated jobs, and cost management.

Takeaways

• Use a VPS for always-on operation to enable schedules and alerts.
• Use Hostinger’s one-click template to minimize setup friction.
• Save and securely store the admin credentials for the Hermes web terminal.
• Enable daily auto-backups and pick an optimal server location.
• Refer to the Hermes agent setup guide for commands and troubleshooting.

---

Summary

This transcript provides a complete beginner-to-intermediate implementation guide for deploying and operating a Hermes AI Agent as a 24/7 autonomous assistant. The core problem addressed is: how to move from a passive chatbot (like ChatGPT) to a persistent, self-improving agent that:

• Runs continuously (via VPS)
• Communicates via messaging platforms (Telegram)
• Learns via skills + memory
• Automates tasks via scheduled jobs
• Optimizes cost via model selection

Primary technologies:

• Hermes Agent
• Docker
• VPS (Hostinger)
• OpenRouter (AI model provider)
• Telegram Bot API

Source:

---

Detailed Step-by-Step Breakdown

1. Deploy Hermes on VPS (Hostinger One-Click)

Recommended VPS Specs:

• 2 vCPU
• 8GB RAM
• 100GB NVMe
• 8TB bandwidth

Steps:

1. Choose VPS plan (KVM-2 recommended)

2. Configure:

   • Billing cycle (recommended: 12 months)
   • Server location (auto)
   • Optional backups

3. Set credentials:

   • Username: hermes
   • Password: auto-generated (SAVE THIS)

4. Deploy → wait for provisioning

---

2. Access VPS Terminal

1. Open Hostinger dashboard
2. Launch Web Terminal
3. Note project ID (last 4 characters)

---

3. Navigate to Hermes Project

cd /docker/hermes-agent-<PROJECT_ID>

---

4. Enter Docker Container

docker compose exec -it hermes-agent /bin/bash

---

5. Run Hermes Setup Wizard

hermes setup

Select:

• Setup mode → Quick Setup
• AI Provider → OpenRouter

---

6. Configure OpenRouter API

Steps:

1. Create account

2. Generate API key

3. Set:

   • Credit limit (e.g., $50)
   • No expiration (optional)

Paste API key into terminal (Note: invisible paste for security)

---

7. Select AI Model

Example:

• Default: Claude Opus (expensive ❌)
• Recommended: GPT-5.4-mini (cheaper)

---

8. Setup Telegram Integration

Create Bot

1. Open Telegram

2. Search: @BotFather

3. Commands:

   • /start
   • /newbot

4. Provide:

   • Bot name
   • Username

5. Copy Bot Token

---

Configure in Hermes

Paste:

• Bot Token
• Allowed User ID

Get User ID:

1. Search: @userinfobot
2. Copy numeric ID

---

Final Setup Prompt

• Confirm home channel → Y

---

9. Launch Hermes Chat

hermes gateway install
hermes gateway run

Result:

• Agent is now:

  • Online 24/7
  • Connected to Telegram
  • Ready for interaction

---

10. First Message (IMPORTANT CONTEXT SETUP)

Send via Telegram:

You're running on a Hostinger VPS inside a Docker container. 
Check your environment and list tools and skills.

---

11. Skills System (Core Learning Mechanism)

Definition:

• Reusable task workflows saved automatically

View skills:

Show me all installed skills

---

12. Create Custom Skill

Example:

Research top trending AI tools. 
Return top 3. 
Then save this as a skill called "YouTube video research"

Result:

• Task executed
• Workflow saved as reusable skill

---

13. Memory System (Persistent Context)

View memory:

Show me what you remember about me

Add preference:

Remember: I prefer short and direct responses

---

14. Upgrade Memory (Optional)

Command:

hermes memory setup

Options:

• Built-in memory (default)
• Honcho (advanced behavioral memory)

---

15. Create Automated Jobs (Scheduling)

Example:

Set a weekly job every Monday at 9AM 
to research top AI tools using YouTube video research skill

Capabilities:

• Cron-like scheduling
• Skill integration
• Automatic execution

---

16. Manage Jobs

Show all current jobs

---

17. Model Switching (Cost Optimization)

Command:

/model

Strategy:

• Cheap model → daily tasks
• Expensive model → complex jobs

---

18. Assign Model per Job

Example:

Change weekly job to use Opus 4.6

---

19. Security Mechanisms

• User ID whitelisting
• Docker container isolation
• Command approval system
• Skill scanning

---

Key Technical Details

Core Commands

• hermes setup
• hermes gateway install
• hermes gateway run
• hermes memory setup
• docker compose exec -it hermes-agent /bin/bash

File Paths

• /docker/hermes-agent-<ID>
• Config + data stored inside container

APIs

• OpenRouter API key required
• Telegram Bot Token required

Models

• GPT-5.4-mini (recommended default)
• Claude Opus (high cost)

Environment

• Docker containerized
• VPS-hosted (24/7 uptime)

---

Pro Tips

• Use cheap models for cron jobs
• Always set spending limits in OpenRouter
• Create skills after every repeated task
• Keep agents running on VPS, not local machine
• Use first message to define environment context
• Start with Telegram only, expand later
• Snapshot VPS before major changes
• Use separate accounts for integrations (security)

---

Potential Limitations/Warnings

• API cost scaling if using expensive models

• Skill quality depends on initial execution

• Memory can become cluttered

• VPS downtime = agent downtime

• Security risk if:

  • API keys exposed
  • Unsafe skills installed

• Telegram bot publicly discoverable (but access-restricted)

---

Recommended Follow-Up Resources

• Docker container basics
• Telegram Bot API docs
• OpenRouter model comparison
• Cron job scheduling concepts
• Prompt engineering for autonomous agents

---

Suggested Books (5)

1. Docker Deep Dive — Nigel Poulton Kindle, Hardcover, Paperback A practical guide to Docker containers, helping you understand how Hermes runs in isolated environments and how to manage, debug, and optimize containerized applications.

2. The DevOps Handbook — Gene Kim et al. Kindle, Paperback Covers automation, infrastructure management, and continuous systems—directly relevant to running a persistent AI agent on a VPS with scheduled jobs.

3. Prompt Engineering for Generative AI — James Phoenix & Mike Taylor Kindle, Paperback Essential for improving how you interact with Hermes, especially when creating high-quality skills and automated workflows.

4. Designing Autonomous AI Agents — Various Authors (O’Reilly) Focuses on multi-agent systems, autonomy, and task orchestration—highly aligned with Hermes' architecture and use cases.

5. Site Reliability Engineering — Google Kindle, Paperback Explains how to maintain reliable always-on systems, which is crucial when running a 24/7 AI assistant on a VPS with real-world dependencies.