FULL Hermes Setup on $15/month in 10 Minutes

📋 Overview

Alejandro AO walks through setting up a fully self-hosted Hermes Agent for roughly $15/month total: ~$5 for a VPS (Hetzner, OVH, or DigitalOcean) plus $10 for a Minimax token plan that provides unlimited-feeling M3 model access with a 1M context window. The twist: instead of manually configuring everything, he uses a Pi coding agent with a dedicated setup skill that automates VPS security hardening, runtime installation, DNS configuration, and the full Hermes setup — you just answer questions when the agent needs input.

1 The $15/month Stack

▶ 0:00

The complete stack breaks down into two components:

• ~$5/month — VPS (Hetzner, OVH, or DigitalOcean) — runs the Hermes Agent with its dashboard, sessions, models, and cron jobs
• $10/month — Minimax Token Plan — provides the LLM that powers the agent, with capped monthly expenses

The result: a self-hosted Hermes instance on your own domain name, with the full web dashboard accessible from anywhere.

2 Choosing a VPS Provider

▶ 0:54

Alejandro compares several budget VPS providers:

• Hetzner — cost-optimized tier at ~€4.79/month (4GB RAM, 2 cores). May have availability issues.
• OVHcloud — ~$6/month for 8GB RAM and 4 cores — actually more powerful than Hetzner's cheapest tier
• DigitalOcean — Droplets are another option at similar price points

The recommendation: pick whichever is available. Any of these providers gives you more than enough resources for Hermes. The $5/month tier is sufficient for most use cases.

3 VPS Setup — SSH Keys & Cloud Config

▶ 1:28

When creating the VPS, two things are needed:

• SSH keys — create your key pair locally, upload the public key to the VPS provider. Alejandro references a separate detailed video for this step.
• Cloud configuration — a cloud-init config that runs on first boot. Alejandro provides his config in the description, linked at his website.

Once the VPS is running, connect via SSH and proceed to install Hermes. The cloud config handles the initial system setup.

4 Minimax Token Plan — $10/month for M3

▶ 3:27

Alejandro's top model recommendation for Hermes is Minimax, specifically their token plan:

• $10/month — the lowest tier, and "absolutely more than enough" for running Hermes
• 1-2 concurrent agents — supports running Hermes and/or OpenCode simultaneously
• M3 model — their latest LLM with a 1 million token context window, natively multimodal
• Token plan concept — like a Codex or Claude Code subscription, but just the tokens — you can plug them into any AI assistant (Hermes, OpenCode, Pi, etc.)

Alejandro shows his actual usage dashboard: after nearly 5 hours of constant use setting up Hermes, he's "nowhere near the quota." The key advantage is capped expenses — you'll never get a surprise bill.

Important: Use the Subscription API Key

When entering the Minimax API key into Hermes, use the one from your subscription, not the pay-as-you-go key. The pay-as-you-go key charges per token with no cap.

5 Hugging Face Inference Providers — The Alternative

▶ 5:33

If you don't want to commit to a $10/month plan, Alejandro suggests Hugging Face Inference Providers as a pay-as-you-go alternative:

• Filter by "inference available" on Hugging Face to find models with direct API access
• No third-party signups — call models directly via Hugging Face's API with just a HF access token
• Model switching — select Hugging Face as an inference provider in Hermes and switch to any available model
• Minimax M2.7 is already available; M3 should drop soon

Creating a Hugging Face Access Token

Go to Hugging Face → Access Tokens → Create new token → enable "Make calls to inference providers." This single token lets you access any inference-available model.

6 AI-Assisted VPS Configuration

▶ 6:28

Here's where it gets interesting: instead of manually configuring the VPS, Alejandro uses a Pi coding agent with a dedicated skill to set everything up.

The workflow:

1. SSH into the VPS
2. Run Pi (already installed via cloud-init)
3. Load the Hermes VPS setup skill
4. Tell Pi: "Help me set up this VPS"
5. Answer questions when the agent needs input (API keys, domain name, etc.)

The agent handles all the terminal work — installing dependencies, configuring services, setting up security — while you just provide the information it asks for.

7 The Hermes VPS Setup Skill

▶ 7:45

The setup skill (available on GitHub) automates the entire Hermes deployment. It covers:

• VPS security hardening
  • Firewall configuration
  • Disable root SSH login
  • Enable fail2ban
  • Standard VPS security best practices
• Runtime dependencies — installs everything Hermes needs to run
• Hermes setup — guides you through the interactive Hermes configuration
• GitHub remote backup — prepares backup integration
• DNS records with Caddy — sets up a reverse proxy so Hermes is accessible on your own domain with HTTPS

Usage: copy-paste the skill into Pi, and it drives the entire installation process.

8 DNS, Domain & Final Configuration

▶ 8:33

The skill also handles setting up Hermes on a custom domain name:

• DNS records — the skill asks you to create DNS records pointing to your VPS IP
• Caddy reverse proxy — handles HTTPS certificate provisioning automatically via Let's Encrypt
• Result — Hermes dashboard accessible at https://your-domain.com with full TLS

The final step: enter the Minimax API key (from the subscription, not pay-as-you-go) and you're running.

🎯 Key Takeaways

🔗 Resources & Links

• Written tutorial — full step-by-step article version
• Hermes VPS Setup Skill — the Pi skill that automates the entire configuration
• Cloud configuration — cloud-init config for VPS first boot
• Hetzner Cloud — budget VPS provider
• OVHcloud VPS — alternative VPS with more specs per dollar
• Minimax — M3 model provider with token subscription plans
• Hugging Face Inference Providers — pay-as-you-go access to open-source models

⏱ Timestamp Index