About Hermes Agent
Hermes Agent is an open source AI agent developed by Nous Research. You use it from the command line, connect it to the model provider of your choice, and it can run around the clock on a server. That makes it more useful on a VPS than on a laptop you shut at the end of the day.
The VPS is delivered with the official installer already run. A hermes system user is created, your root SSH keys are copied to it so you can log in directly with that account, and the hermes command is installed along with its Python dependencies (through uv). Git, curl and sudo are present.
One step is left for you: configuration. It requires choosing a model provider and entering an API key, so it is interactive and does not run automatically. Once you are logged in, run hermes setup.
What you can do with Hermes Agent
Official installer already run
The hermes command is installed for the hermes user with uv and the Python dependencies, using the script published by Nous Research. There is nothing to compile and no environment to manage.
Dedicated account with your SSH keys
The hermes account inherits root's authorized keys. Its password is locked, so login works by key only.
Guided setup with hermes setup
The interactive assistant asks for your model provider and API key. You keep the choice of model and can change it later.
Persistent user services
systemd linger is enabled for hermes. Services running under its account, such as the gateway, stay active with no open session.
Sudo rights for automation
The hermes user has passwordless sudo, handy for installing tools the agent needs. Restrict it to suit your use.
Debian or Ubuntu, your choice
The recipe works on Debian 12 and 13 as well as Ubuntu 20.04, 22.04 and 24.04, selectable at order time.
Get started with Hermes Agent
From order to production in a few minutes.
- 1
Order the VPS with Hermes Agent
Choose a Linux VPS in Paris or Frankfurt, select Debian or Ubuntu, then tick Hermes Agent in the Applications tab. Add your public SSH key to the order: it will be copied to the hermes account.
- 2
Connect over SSH as the hermes user
Once the server is delivered, let cloud-init finish the installation, then open a session as hermes. If the hermes command is not found, open a new session to reload the environment.
Bashssh hermes@SERVER_IP cloud-init status --wait hermes --help - 3
Configure the model provider
Start the assistant and answer the prompts: provider, API key, model. Have the API key for your chosen provider ready before you begin.
Bashhermes setup - 4
Secure the server for production
Restrict network access to SSH only with a firewall and keep key-based authentication. Schedule a backup of the hermes home directory, which holds your configuration and API keys.
Bashsudo apt install -y ufw sudo ufw allow OpenSSH sudo ufw enable
Hermes Agent use cases
Always-on AI assistant
Keep an agent running on a server that never gets switched off, reachable over SSH from any machine. Your working sessions no longer depend on your laptop.
Gateway to your messaging tools
The persistent user service lets you run the Hermes Agent gateway in the background. The agent stays reachable without an SSH session being open.
Sysadmin task automation
With git, curl and sudo available, the agent can draft scripts, inspect repositories or install tools on the machine. It is a good testing ground on a server kept separate from production.
Sandbox for testing models
Compare several model providers with the same agent by repeating the configuration. These experiments stay isolated from your personal environment.
Hermes Agent FAQ
Configuration needs a model provider and an API key, which only you have. The recipe therefore installs the agent and leaves you to run hermes setup as the hermes user.
More about Hermes Agent
Check that the installation has finished
The installation runs on first boot through cloud-init. It downloads uv and the Python dependencies, which takes a few minutes. To wait for it cleanly:
Bashcloud-init status --wait
The final message tells you to log in as hermes and run hermes setup. If you prefer to follow the details, read /var/log/cloud-init-output.log.
Run the setup
hermes setup is interactive: it asks for your model provider and API key. The recipe does not run it for you, because it cannot guess that information. You can run the assistant again later to switch providers.
Background services
The recipe enables loginctl enable-linger hermes. Without it, systemd stops a user's services when their last session closes. With linger, a user service such as the Hermes Agent gateway keeps running after you disconnect from SSH.
Security
The hermes account has a locked password and passwordless sudo. Only your SSH key opens the door, but anyone who obtains it has root rights. Protect that key with a passphrase.
An agent that runs commands can also make mistakes. If you let it act on the server, remove the passwordless sudo once your tools are installed (the file is in /etc/sudoers.d/), and keep no critical data on this machine.
Backups
API keys and configuration live in the user's home directory, /home/hermes. Back it up regularly, for example with rsync to another server, and keep your API keys in a secrets manager of your own. For broader Linux setups, see our Linux VPS and our tutorials.
