Guide

Hermes Agent 24/7 Hosting

Hermes only feels like an assistant when it answers at 3am and runs its scheduled jobs while you sleep. Here is what has to stay running, the commands that keep it up, and two defaults that quietly lose messages or jobs when it is not.

The short answer

To run Hermes Agent 24/7, run its messaging gateway as a service on a machine that never sleeps. On Linux that is hermes gateway install plus loginctl enable-linger; in Docker it is a restart policy; with managed hosting it is done for you. Then fix the one default that bites always-on setups: Telegram messages sent while the gateway was down are dropped on its next start unless you turn that off.

What actually needs to run 24/7

The piece that matters is the gateway: the long-running Hermes process that connects to Telegram, Discord, Slack, WhatsApp and the other platforms. It receives your messages, runs the agent, and hosts the cron scheduler, which checks for due jobs every minute. Running hermes in a terminal for a chat is not enough; when that terminal closes, nothing is listening.

The AI model itself runs at your provider, so an always-on Hermes host can be small. The Hermes Docker guide lists 1 GB of RAM as the minimum and 2 to 4 GB as the recommendation.

Your options

Where it runs	Always on?	Your work
Laptop	Only while awake and online	Low, but it stops when the lid closes or the machine sleeps
Home server, mini PC or NAS	Yes, if it never sleeps	Service setup, updates, power and network outages
VPS	Yes	Install, service, updates, monitoring, security
Managed Hermes hosting	Yes	None for the server; you just use the agent

Linux or VPS: install the gateway as a service

hermes gateway install            # user systemd service
sudo loginctl enable-linger $USER # start at boot, survive logout
hermes gateway start
hermes gateway status
journalctl --user -u hermes-gateway -f   # follow the logs

Without enable-linger, a user service stops when you log out, which is exactly what happens when your SSH session ends. On a VPS you can instead install a boot-time system service that still runs as your user with sudo hermes gateway install --system. The Hermes docs point out that a user service with linger lets hermes update restart the gateway without root, so it is the lower-friction choice for a headless box. The installed unit restarts the gateway automatically if it exits. Avoid having both the user and the system unit installed at once.

To restart after a config change, prefer hermes gateway restart: it lets in-flight replies finish first, where a plain systemctl restart does not wait.

macOS: a launchd agent

hermes gateway install writes ~/Library/LaunchAgents/ai.hermes.gateway.plist, which starts the gateway when you log in and logs to ~/.hermes/logs/gateway.log. The plist captures your shell PATH at install time, so run hermes gateway install again after installing new tools such as a new Node.js version. And a Mac that sleeps is a gateway that is offline: an always-on Mac needs its sleep settings changed, not just a service.

Docker

The official image keeps your data in /opt/data and supervises the gateway inside the container, restarting it within seconds after a crash. Give the container a restart policy so it also comes back after a reboot; the Hermes docs use --restart unless-stopped. Never point two gateway containers at the same data directory. Our Hermes Agent Docker guide walks through the setup.

The two defaults that lose messages and jobs

1. Telegram drops messages sent while the gateway was offline

On a cold start, Hermes asks Telegram to discard the updates that queued up while it was down. The Hermes docs describe this as fitting always-on servers, where a restart means clean up. On any host that ever switches off, it means messages people sent overnight vanish, with no log and no retry. Turn it off:

platforms:
  telegram:
    extra:
      drop_pending_on_cold_boot: false

Discord has the opposite default: messages missed during an outage are not replayed unless you enable the opt-in discord.missed_message_backfill option for the channels you care about.

2. Missed cron runs collapse into one

When the gateway comes back after downtime, a recurring job that came due while it was down runs once to catch up, no matter how many runs were missed. That is usually what you want for a daily report. If a late run would do harm, for example a time-sensitive alert, set cron.catch_up_missed: false, which skips a run once it is later than its grace window (half the job's period, between 2 minutes and 2 hours). hermes cron doctor shows jobs that ran late or were caught up.

Reliability checklist

The gateway runs as a service (systemd, launchd or a Docker restart policy), not in a terminal.
Linger is enabled, or a system service is installed, so logout and reboot do not stop it.
The host never sleeps.
drop_pending_on_cold_boot: false is set if the host can ever go offline.
You have decided whether missed cron runs should catch up.
You send the agent a real message after every update or restart, not just a status check.
Someone is told when it goes down, because a quiet agent looks exactly like a lazy one.

Or skip the server: managed Hermes hosting

On OpenClaw Launch the gateway runs on our servers, with restarts, updates and monitoring handled for you, and it is live in under a minute. You connect Telegram, Discord, Slack, WhatsApp or WeChat from the dashboard instead of editing config files. Lite is $3 for the first month, then $6 a month, with 1 vCPU, 2 GB of RAM and $1 a month of AI credits. Pro is $20 a month with 2 vCPU, 4 GB of RAM and $10 a month of credits. There is a free 30-minute trial with no credit card.

One thing to know: a bot that sees no activity at all for 7 days is put to sleep to free capacity, and comes back when you press Wake Up on its dashboard card. A bot you actually use keeps running.

Frequently Asked Questions

Can I run Hermes Agent 24/7 on my laptop?

Only while the laptop is awake. The Hermes gateway is what answers your messages and runs scheduled jobs, and it stops whenever the machine sleeps, shuts down or loses its connection. For a true always-on agent, run it on a machine that never sleeps, a VPS, or managed hosting.

What command keeps Hermes running after I log out?

On Linux, run hermes gateway install to create a user systemd service, then sudo loginctl enable-linger $USER once so it starts at boot and survives logout. On a VPS you can instead install a boot-time system service with sudo hermes gateway install --system. On macOS, hermes gateway install creates a launchd agent that starts the gateway at login.

Do I lose Telegram messages sent while Hermes was offline?

By default, yes. On a cold start the Hermes Telegram adapter drops the updates Telegram queued while the gateway was down, silently and without a log line. If your host ever switches off, set platforms.telegram.extra.drop_pending_on_cold_boot to false so that backlog is delivered in order when the gateway comes back.

What happens to cron jobs when Hermes is down?

The scheduler runs inside the gateway. When it comes back, a recurring job that came due while it was down runs once to catch up, however many runs it missed, and paused jobs never catch up. Set cron.catch_up_missed to false to skip a run that is later than its grace window (half its period, between 2 minutes and 2 hours); a run inside that window still fires late.

Do I need Docker to run Hermes 24/7?

No. A systemd or launchd service is enough. Docker is a good alternative on a server: the official image supervises the gateway and restarts it within seconds if it crashes, and the Hermes docs run it with --restart unless-stopped. Never run two gateway containers on the same data directory.

Does a managed Hermes bot on OpenClaw Launch ever sleep?

A bot you use runs around the clock. If a bot sees no activity at all for 7 days, no chats, no dashboard visits and no credit use, it is put to sleep to free server capacity, and you bring it back with the Wake Up button on its dashboard card. If you rely only on scheduled jobs with nobody chatting, check in on the bot from time to time.

Related Guides

Managed Hermes Agent Hosting — plans and what is included
Best VPS for Hermes Agent — if you want your own server
Hermes Agent in Docker — container setup
Hermes Agent Team Hosting — one agent for a whole team
Hermes Agent System Requirements — hardware in detail
Run an AI Agent 24/7 — the general version of this guide