Hermes Agent VPS Setup Guide: Ubuntu, Security & 24/7 Gateway

This guide shows how to install Hermes Agent on an Ubuntu VPS, connect a model, test the CLI, configure messaging, and keep the gateway running after you disconnect from SSH.

Recommended starting server: 1 vCPU, 2 GB RAM, 40 GB SSD, and Ubuntu 24.04 LTS. Choose 4 GB RAM if you plan to use browser automation, several channels, or concurrent tasks. A GPU is unnecessary when Hermes uses a hosted model API.

Fast path: Create the VPS, log in with SSH, run the official installer as a non-root user, configure a model with hermes model, verify a normal chat, and only then install the messaging gateway as a service.

Last checked: July 8, 2026. Commands were verified against the official Hermes quickstart and messaging gateway guide.

What You Need

• An Ubuntu 22.04 or 24.04 VPS
• Root or sudo access
• An SSH key or initial root password
• An API key or supported OAuth login for your model provider
• A model with at least a 64K context window
• Optional bot tokens for Telegram, Discord, Slack, or other channels

If you are still deciding whether Hermes is the right agent, read our Hermes Agent review first.

Hermes Agent VPS Requirements

Workload	vCPU	RAM	Storage	Recommendation
CLI testing with a hosted model	1	1-2 GB	25 GB	2 GB avoids tight memory during updates
Always-on gateway, 1-2 channels	1	2 GB	40-50 GB	Best low-cost starting point
Browser tools or several workflows	2	4 GB	50-80 GB	Recommended for regular use
Multiple agents or heavy automation	4+	8 GB+	80 GB+	Monitor and size for actual load

These figures are for API-based inference. Running a local model is a different workload and usually needs substantially more RAM, disk, and possibly GPU capacity.

Quick Option: Pre-Installed Hermes Agent VPS

If you do not want to install the runtime manually, LightNode's Hermes Agent VPS provides a pre-installed image. Its entry configuration was listed at 1 vCPU, 2 GB RAM, 50 GB NVMe storage, and $7.71 monthly equivalent when checked on July 8, 2026.

This option is useful when deployment speed and hourly billing matter. A normal Ubuntu VPS offers more control and is better for learning how the system is configured.

Affiliate disclosure: The LightNode link is an affiliate link. We may earn a commission at no extra cost to you. This does not change the setup steps or selection criteria in this guide.

Compare it with other options in our best Hermes Agent VPS hosting guide.

Step 1: Create the Ubuntu VPS

Select:

• Ubuntu 24.04 LTS, or 22.04 LTS if 24.04 is unavailable;
• a region near you and the services Hermes will call;
• at least 2 GB RAM for a comfortable first deployment;
• SSH key authentication when the provider supports it.

Copy the public IP address after deployment.

Step 2: Connect with SSH

On macOS, Linux, or Windows Terminal:

ssh root@YOUR_SERVER_IP

Replace YOUR_SERVER_IP with the address from the VPS dashboard.

Confirm the operating system and resources:

cat /etc/os-release
free -h
df -h

Step 3: Update Ubuntu

apt update
apt upgrade -y
apt install -y curl ca-certificates ufw

Reboot if Ubuntu reports that a restart is required:

reboot

Wait a minute, then reconnect.

Step 4: Create a Non-Root User

Do not run a tool-enabled agent as root for daily use.

adduser hermes
usermod -aG sudo hermes

If root already uses SSH-key authentication, copy the authorized key to the new account:

install -d -m 700 -o hermes -g hermes /home/hermes/.ssh
cp /root/.ssh/authorized_keys /home/hermes/.ssh/authorized_keys
chown hermes:hermes /home/hermes/.ssh/authorized_keys
chmod 600 /home/hermes/.ssh/authorized_keys

Open a second terminal and verify the new login before closing the root session:

ssh hermes@YOUR_SERVER_IP

Step 5: Enable a Basic Firewall

Allow SSH before enabling UFW:

sudo ufw allow OpenSSH
sudo ufw enable
sudo ufw status verbose

Hermes messaging integrations usually make outbound connections, so you should not open arbitrary inbound ports. Only add a firewall rule after a specific integration requires it and you understand its authentication model.

Step 6: Install Hermes Agent

Run the current official installer while logged in as the hermes user:

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

Reload the shell:

source ~/.bashrc

Verify the installation:

hermes doctor
hermes --help

The installer manages the required Hermes runtime and dependencies. Avoid mixing this installation with random third-party scripts unless you have reviewed them.

Step 7: Configure a Model Provider

Run:

hermes model

Choose a supported provider and complete the authentication prompts. Hermes can use Nous Portal, OpenAI, Anthropic, OpenRouter, Gemini, local OpenAI-compatible endpoints, and many other providers.

For the guided all-in-one flow, use:

hermes setup

Hermes separates configuration and secrets:

• non-secret settings: ~/.hermes/config.yaml
• API keys and tokens: ~/.hermes/.env

Protect both files, never commit them to Git, and never paste their contents into public issue reports.

Step 8: Test the CLI Before the Gateway

Start Hermes:

hermes

Use a simple, verifiable prompt:

Report the current hostname, operating system, free memory, and disk usage. Do not change anything.

Confirm that:

• the expected model appears in the startup banner;
• Hermes replies normally;
• read-only terminal use works;
• a second message continues the same session.

If basic chat fails, do not add messaging or scheduled tasks yet. Run:

hermes doctor
hermes model

Fix provider authentication first.

Step 9: Configure Telegram, Discord, or Another Channel

Start the interactive gateway wizard:

hermes gateway setup

The wizard supports Telegram, Discord, Slack, WhatsApp, Signal, Email, Home Assistant, Microsoft Teams, and other platforms. It prompts for the relevant bot token or credentials.

Use pairing or an explicit allowlist. An internet stranger should never be able to message an agent with terminal and file access.

Test the gateway in the foreground first:

hermes gateway

Send a message from your configured platform. Stop the foreground process with Ctrl+C after the test succeeds.

Step 10: Keep Hermes Running 24/7

Install the gateway as a user service:

hermes gateway install
hermes gateway start
hermes gateway status

On a systemd-based VPS, enable lingering so the user's service can run without an active SSH session:

sudo loginctl enable-linger hermes

The official CLI also supports a system-wide Linux service:

sudo hermes gateway install --system
hermes gateway status --system

Use one service mode for an instance, not both. Prefer the user service unless you have a specific reason to run a system service.

Step 11: Apply Safe Tool Defaults

Hermes can execute commands, so server isolation matters more than a polished dashboard.

For an internet-facing or unattended gateway:

• keep the default manual approval mode for dangerous commands;
• do not launch sessions with --yolo;
• restrict who can message the agent;
• separate valuable production credentials from the agent account;
• consider the Docker terminal backend for command isolation;
• disable tools the workflow does not need.

Configure tool access interactively:

hermes tools

To use Docker as the terminal boundary after Docker is correctly installed:

hermes config set terminal.backend docker

Read the official Hermes security guide before enabling unattended automation.

Step 12: Update and Back Up Hermes

Check the installation:

hermes doctor
hermes gateway status

Update Hermes:

hermes update
hermes gateway stop
hermes gateway start

Create a private local backup before major changes:

umask 077
tar -czf "$HOME/hermes-backup-$(date +%F).tar.gz" "$HOME/.hermes"

The archive may contain API keys, bot tokens, memory, and conversation data. Encrypt it before copying it to remote storage, and test restoration on a disposable server.

Keep Ubuntu patched:

sudo apt update
sudo apt upgrade -y

Troubleshooting

hermes: command not found

Reload the profile or reconnect:

source ~/.bashrc

Then inspect the installer's final output for the installed binary path.

Hermes starts but the model does not respond

Run:

hermes doctor
hermes model

Recheck the API key, model name, provider endpoint, billing status, and minimum 64K model context.

The bot does not reply

hermes gateway status
hermes gateway setup

Confirm the bot token, user pairing or allowlist, and whether another process is already using the same bot credentials.

The gateway stops after logout

Confirm the service is installed, then enable user lingering:

hermes gateway install
sudo loginctl enable-linger hermes
hermes gateway start

The VPS runs out of memory

Check usage:

free -h
ps aux --sort=-%mem | head

Browser automation and concurrent tasks can exceed a 2 GB plan. Upgrade to 4 GB rather than treating swap as a permanent fix.

Common Setup Mistakes

• Installing as root: this gives agent-generated commands unnecessary control over the server.
• Opening a dashboard port to the public internet: use authenticated private access or an SSH tunnel instead.
• Adding every tool immediately: begin with the minimum toolset and expand only for a real workflow.
• Skipping the CLI test: a broken provider configuration becomes harder to diagnose after messaging and services are added.
• Ignoring model costs: the VPS may be the smaller part of the monthly bill.
• Backing up without protecting secrets: ~/.hermes/ can contain sensitive credentials and personal context.

Manual VPS or Pre-Installed VPS?

Choose manual installation when…	Choose pre-installed hosting when…
You want full control of the OS and runtime	You want the shortest path to a working server
You are comfortable maintaining Ubuntu	You prefer an image with Hermes already installed
You want to audit each installation step	You are testing and value hourly billing
You need a custom security or networking layout	Standard configuration is enough

If you choose the manual route, almost any reliable Ubuntu VPS with adequate RAM will work. The provider matters less than correct permissions, backups, and monitoring.

FAQ

What is a Hermes Agent VPS?

It is a virtual Linux server that runs Hermes Agent continuously, even when your laptop is off. It hosts the gateway, memory, skills, and scheduled tasks; a hosted LLM usually performs the inference.

What is the minimum RAM for Hermes Agent?

One GB can work for a minimal API-based CLI setup, but 2 GB is the practical starting point for a VPS. Use 4 GB for browser automation or several simultaneous workflows.

Does Hermes Agent require a GPU?

No. A normal CPU VPS is sufficient when you use OpenAI, Anthropic, OpenRouter, Nous Portal, or another hosted model provider.

Can I run Hermes Agent on a $5 VPS?

Yes, if the plan has enough RAM and you keep the workload light. Compare the full monthly cost, including IPv4, backups, traffic, tax, and model API usage.

Which Ubuntu version should I use?

Ubuntu 24.04 LTS is the preferred new-server choice. Ubuntu 22.04 LTS is also suitable.

How do I keep Hermes Agent running after SSH closes?

Install the gateway service with hermes gateway install, start it with hermes gateway start, and enable systemd lingering for the hermes user.

Where does Hermes Agent store its data?

The default data directory is ~/.hermes/. It can include configuration, secrets, sessions, memories, skills, pairing data, and runtime files.

Is it safe to expose Hermes Agent publicly?

Do not expose unauthenticated agent interfaces. Use pairing or allowlists for messaging, authenticated private access for web interfaces, manual command approvals, and container isolation where appropriate.

Final Checklist

• Ubuntu is updated
• Hermes runs under a non-root user
• SSH works for the new user before root access is restricted
• UFW allows SSH and no unnecessary ports
• hermes doctor passes
• A normal CLI conversation works
• Messaging users are paired or allowlisted
• The gateway service survives logout and reboot
• Dangerous commands require approval
• ~/.hermes/ is backed up securely

Once these checks pass, the VPS is ready for a real always-on Hermes workflow rather than just a successful installation.