What is the minimum VPS spec for running Hermes 24/7?

For cloud LLM backends (API-based), 1GB RAM and 1 vCPU handle Hermes comfortably. Hetzner CX11 at 3.29/month is the community standard. For running local Ollama models, you'll need 4GB+ RAM minimum for 7B–13B models, and 16GB+ for 27B models or a GPU instance.

How do I keep Hermes running automatically after a VPS reboot?

Run `hermes gateway enable-autostart` on Linux — this creates a systemd service that restarts Hermes automatically after reboot. Verify it worked with `systemctl status hermes-gateway`. This is the most reliable approach for headless VPS deployments.

What data should I back up from a Hermes VPS?

Back up the ~/.hermes/ directory — specifically ~/.hermes/memories/ (MEMORY.md, USER.md), ~/.hermes/skills/ (your custom skills), and ~/.hermes/config.yaml (configuration). Use git for version control on skills and rsync or cloud sync for offsite backup of memory files.

How do I reduce VPS costs when I'm not actively using Hermes?

Use Daytona or Modal integration — both hibernate the VPS when idle and wake on a gateway trigger (like an incoming Telegram message), keeping costs near zero between uses. Ideal for personal use where you only check in a few times per day.

VPS Hosting Guide

Q: Can I run multiple Hermes instances on one VPS?

Yes. Use Hermes profiles: `hermes profile create work` and `hermes profile create personal` create isolated instances with separate memory, skills, config, and gateway services from the same installation. Each profile has its own conversation history and skill library.

Running Hermes on a VPS gives you an always-on agent — unlike ChatGPT (cloud-only) or Copilot (IDE-only) accessible from anywhere. Here are the best options and how to set up.

VPS Options Compared

Provider	Plan	Monthly	RAM	Notes
Hetzner	CX11	€3.29	2GB	Community favorite
DigitalOcean	Basic	$6	1GB	Simple UI
Linode	Nano	$5	1GB	Now Akamai
Linode	Premium	$24	4GB	For local models
Lambda Labs	GPU	$40-80	16-80GB	For GPU workloads

Hetzner CX11 at €3.29/mo is the community consensus for cloud LLM backends.

Minimum Specs

Cloud LLM (API-based):

1GB RAM
1 vCPU
20GB SSD storage

Local models (Ollama 7B-13B):

4GB RAM minimum
More capable models need 8GB+

Local models (27B+):

16GB+ RAM
GPU recommended
$40-80/mo GPU instances

Why Hetzner

Price: €3.29/mo for 2GB — cheapest practical option
Reliable: Good uptime, decent support
Locations: EU and US datacenters
Community: Well-documented in Hermes Discord

Setup on VPS

ssh root@your-server-ip
# Install Hermes from docs
hermes model setup  # configure your provider
hermes gateway install  # set up as service
hermes gateway start  # start the gateway

For persistent storage across reboots, ensure your data directories persist or bind mount storage.

SSH Access

Your VPS runs Hermes headless. Access via:

SSH + CLI when connected
Telegram/Discord gateway for mobile
The TUI with SSH tunneling

Security Considerations

Run the gateway as a non-root user
Configure firewall (ufw allow for SSH, Telegram port)
Use API keys stored in environment, not config files

Consider Docker sandboxing:

hermes config set terminal.backend docker

Backup

Your Hermes data lives in ~/.hermes/. Back up:

~/.hermes/memories/ — MEMORY.md, USER.md
~/.hermes/skills/ — custom skills
~/.hermes/config.yaml — configuration

Use git for version control, rsync or a cloud sync for offsite backup.

Idle Cost Reduction

For infrequent use:

Daytona integration: Hibernates when idle, wakes on gateway trigger
Modal integration: Serverless GPU, wakes on demand

Both significantly reduce idle costs.

Cron scheduling — automate tasks 24/7

Setup guide Docker install

FAQ

Can I run multiple Hermes instances on one VPS? Yes — use profiles: hermes profile create [name]

Deploy your VPS now

flyhermes.ai

Host Hermes on a VPS