Deploy to a server — $5 for 24/7 uptime

So far, Hermes runs on your own computer. Close the lid, and it stops. This step moves it to a server so it becomes a 24-hour online, phone-reachable personal AI assistant.

How much does it cost

The answer might surprise you: $5 a month is all you need.

Without a local LLM, Hermes uses less than 500MB of memory. The cheapest VPS handles it easily. Model inference goes through an API (OpenRouter, Nous Portal, etc.), and light usage runs $2-5/month in API fees.

Total: less than a fancy cup of coffee per month for an AI assistant with memory, real capabilities, and 24/7 availability.

Pick a VPS

Three solid options:

Provider	Monthly cost	Notes
Hetzner CX22	~$4	Best value, European nodes
DigitalOcean Droplet	$5	Singapore / US West nodes
Vultr	$5	Tokyo node, low latency

Any of them work. Pick Ubuntu 22.04 LTS as your OS.

Step 1: SSH in

After purchasing the VPS, connect via SSH:

ssh root@your_server_ip

Step 2: Install Docker

curl -fsSL https://get.docker.com | sh

Wait a minute or two. Verify:

docker --version

Step 3: Run Hermes

docker pull nousresearch/hermes-agent:latest

Before running, create the config file on the server:

mkdir -p ~/.hermes
nano ~/.hermes/config.yaml

Fill in a minimal config:

model:
  provider: openrouter
  api_key: sk-or-xxxxx        # Your OpenRouter API key
  model: anthropic/claude-sonnet-4

If you use Nous Portal or another provider, adjust according to Step 6.

Save, then start:

docker run -d \
  --name hermes \
  --restart unless-stopped \
  -v ~/.hermes:/opt/data \
  nousresearch/hermes-agent:latest

Key flags:

-d: Run in background
--restart unless-stopped: Auto-restart on crash or server reboot
-v ~/.hermes:/opt/data: Mount your config, memory, and Skills into the container. Delete and rebuild the container anytime — your data stays

Checkpoint — Run docker logs hermes and check the output. If you see Hermes starting up successfully, the server side is done.

Step 4: Connect Telegram

If you created a Telegram bot in Step 11, add the token to the server config.

Edit ~/.hermes/.env on the server:

nano ~/.hermes/.env

Add:

TELEGRAM_BOT_TOKEN=your_bot_token
TELEGRAM_ALLOWED_USERS=your_telegram_user_id

Restart the container:

docker restart hermes

Open Telegram on your phone and send your bot a message. If it replies — congratulations, you now have a 24/7 AI assistant.

Checkpoint — Send a message from your phone, bot replies. Close your laptop, send another message, bot still replies. That's the difference between "running locally" and "deployed."

Security tips

Your Hermes is running on a public server. A few things to keep in mind:

Use Docker, not bare metal. Set terminal: docker in your config so Hermes executes shell commands inside a container, not on the host filesystem.

Restrict Telegram users. Always set TELEGRAM_ALLOWED_USERS with only your own user ID. Without it, anyone who discovers your bot's username can talk to it.

Protect your API keys. ~/.hermes/.env and config.yaml contain your keys. Make sure the server uses SSH key authentication and password login is disabled.

Updating Hermes

docker pull nousresearch/hermes-agent:latest
docker stop hermes && docker rm hermes
docker run -d \
  --name hermes \
  --restart unless-stopped \
  -v ~/.hermes:/opt/data \
  nousresearch/hermes-agent:latest

Because all data lives in ~/.hermes/, deleting and rebuilding the container doesn't affect anything — memory, Skills, and config are all intact.

Even cheaper options

If $5/month still feels like too much, Hermes also supports serverless backends:

Daytona: The environment hibernates when idle, wakes up when a message arrives. Near-zero cost between sessions.
Modal: Same hibernation model, good for light usage.

Set terminal: daytona or terminal: modal in config.yaml.

What you have now

At this point, you have:

A 24/7 online AI agent
Reachable from your phone anytime (Telegram)
Persistent memory that deepens the more you use it
A Skill system that gets more precise over time
Scheduled tasks that work proactively
All data on your own server

Monthly cost: less than a cup of coffee. This is Hermes in its complete form.