How to run Hermes Agent 24/7: VPS or hosted

Short answer. Hermes Agent answers Telegram and Discord and runs its scheduled jobs only while its gateway process runs, so a 24/7 agent needs a machine that never sleeps. On your own VPS: a Linux server with at least 2 GB of RAM and Nous Research's install script run as a normal user over SSH (it starts the setup wizard and installs the gateway as a service), then sudo loginctl enable-linger $USER so the service starts at boot; you run hermes update and keep the server safe. Or rent a server that runs Hermes Agent for you, where the panel does the install, updates and restarts and you only paste a key and a bot token.

Checked on 7 November 2026, accurate for Hermes Agent v0.23.0.

This guide uses Hermes Agent's own documentation (hermes-agent.nousresearch.com/docs) and what we measured running v0.23.0 on our own servers. To go straight to the hosted route, see Option 3.

Why a laptop is not enough

Hermes Agent's messaging gateway is one long-running process that serves every chat app and its scheduler. When the laptop sleeps, reboots or loses its network, the agent stops answering and scheduled jobs do not run. For a headless server, Hermes Agent's docs describe a service that starts at boot and survives logout.

What you need

You also need an API key from an AI provider (see which AI provider key to use) and a chat app to talk to it, such as a Telegram or Discord bot. More on memory in how much memory Hermes Agent needs.

Option 1: your own VPS with the install script

1. Create the server with Ubuntu and at least 2 GB of RAM, and connect over SSH. Hermes Agent's Docker page warns that some providers' browser consoles mangle pasted characters such as : and @; SSH avoids that.
2. Use a normal user, not root. Hermes Agent's docs say to run the installer as the user that will run the service. The agent's terminal tool runs commands on this machine as that user, so keep that account for the agent.
3. Install. The installer installs the latest stable release:

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

   To keep the install smaller, its --skip-browser option leaves out its Chromium, so local browser automation is off until you run hermes pm install chromium.
4. Configure. Run over SSH, the installer starts the setup wizard itself: it asks for your AI key, your bot tokens and the user IDs allowed to message the agent. See Telegram and Discord. Later you can run it again with hermes setup, or step by step with hermes model and hermes gateway setup.
5. Run it as a service. At the end the installer installs the gateway as a user service if setup did not already; if you skipped that, run hermes gateway install. Then, once, let it start at boot and keep running after you log out:

   sudo loginctl enable-linger $USER
   hermes gateway status
   journalctl --user -u hermes-gateway -f

   Hermes Agent's docs recommend this user service with lingering for a headless server: it starts at boot, and updates can restart it without root. The unit Hermes Agent writes restarts the gateway when it stops unexpectedly and after an in-chat /restart. The docs also describe a system-wide service (hermes gateway install --system), but then every restart needs root, and they warn against keeping both kinds installed.
6. Update with hermes update. It follows stable releases, pauses the service, updates and starts it again. Releases are frequent (v0.22.5 on 23 October, v0.22.6 on 2 November, v0.22.7 on 3 November and v0.23.0 on 5 November 2026), and in our test an update needed more memory than the install.

Option 2: your own server with Docker

Hermes Agent publishes an official image, nousresearch/hermes-agent, with all data in one folder you mount at /opt/data. Run the setup wizard once, then the gateway in the background:

mkdir -p ~/.hermes
docker run -it --rm -v ~/.hermes:/opt/data nousresearch/hermes-agent setup
docker run -d --name hermes --restart unless-stopped -v ~/.hermes:/opt/data nousresearch/hermes-agent gateway run

The docs also publish port 8642, for Hermes Agent's API server and health check; they say it is optional if you only use chat apps such as Telegram and Discord. Docker installs do not support hermes update: you update by pulling a new image and recreating the container. Inside the image a process supervisor restarts the gateway if it crashes.

Keep a self-hosted agent safe

• Allow only yourself. With no allowlist Hermes Agent denies everyone, which is safe. Never set GATEWAY_ALLOW_ALL_USERS=true on an agent that can run commands.
• Keep the dashboard off the internet. Hermes Agent's docs say its username and password sign-in is "not suitable for direct public-internet exposure"; for public deployments they offer Nous Portal or OpenID Connect sign-in. On your own VPS, the simplest is to reach it through an SSH tunnel.
• Leave the API server off unless you need it. It is off by default and listens on 127.0.0.1 when on.
• Know what the agent can touch. With the default local terminal, commands run on the server itself, as the user that runs Hermes Agent. It checks commands against a list of dangerous patterns and, in its default "smart" mode, asks a model to judge them; uncertain ones come to you in the chat for approval. Hermes Agent's production checklist recommends running its commands in a container instead (terminal.backend: docker in config.yaml).
• Back up ~/.hermes: it holds the settings, keys, memory, skills and sessions.

Option 3: a server that runs it for you

A PlayNHost Hermes Agent server does the steps above for you:

• Install: the first start runs Nous Research's own installer, without the browser tools; in our tests it took about 2.5 to 4 minutes.
• Updates: every start checks for a new stable release and installs it first (about 20 seconds when there is nothing new). If an update fails, the version you have starts as before. You can turn this off with Update Hermes Agent on every start under Code, Settings.
• Restarts: an in-chat restart or update is started again for you, and if the agent stops by itself the panel starts it again within about a minute. Under Restarts you can also schedule a daily restart.
• Safety: only the user IDs on the Agent page can message the agent; nothing on the server listens on the internet; WhatsApp is refused; the agent is stopped if it opens a port or lets someone else in.
• Your part: paste one AI provider key and a Telegram or Discord bot token on the Agent tab, add your user ID and press Start. Your keys and tokens are stored encrypted and never shown again.

What you give up compared with your own VPS: there is no command line (no hermes commands, no Nous Portal sign-in, no pairing approvals), no web dashboard, no browser automation, only Telegram and Discord, and one agent per server (Hermes Agent profiles are refused). You can still edit the agent's files in Files, and its log is .hermes/logs/gateway.log.

VPS or hosted: which to pick

Our judgment: pick a VPS if you want every Hermes Agent feature or already run servers; pick hosted if you want a Telegram or Discord agent without the upkeep. Hermes Agent itself is free and MIT-licensed either way; Nous Research's docs warn that nobody legitimate asks you to pay for the agent itself. A host charges for the server, not the agent. Next: create a Hermes Agent server.

PlayNHost is not affiliated with Nous Research. Hermes Agent is open-source software under the MIT licence.