Skip to content
vavio

Does closing my terminal stop Hermes and its scheduled jobs?

In Hermes v0.23.1, closing the normal chat terminal detaches your view; accepted work belongs to the gateway and is not cancelled by that detach. Closing a terminal that is running the gateway itself in the foreground can stop that gateway. Scheduled jobs need a live scheduler and, for agent-backed jobs, a live gateway on a computer that stays awake.

That distinction matters more than whether the terminal window is open. This answer follows official source at commit 3b42157, checked on 14 November 2026. It is a source review, not a test of closing a terminal or running a scheduled job. Older Hermes releases and custom launchers can behave differently.

A chat window and the gateway have different lifetimes

The normal hermes chat client discovers or starts the profile’s gateway, then attaches to a session there. In that client, /quit, /exit, /detach, end-of-input and closing the terminal detach the view. /stop explicitly interrupts execution. Save the printed session ID if you want to resume the same conversation:

hermes --cli chat --resume SESSION_ID

Replace SESSION_ID with your actual ID. Reopening a session lets you inspect its result or pending request; it does not establish that the task finished while you were away. Approvals, credentials, model limits and errors can still block it.

An automatically started gateway is also different from an installed background service. In this release, an unmanaged client-started gateway can exit after ten idle minutes by default. Its idle check keeps it up for runnable scheduled jobs, active work and other duties. That convenience does not install a service that starts after a reboot.

What you close or stop What to expect
Normal gateway-backed chat client The view detaches; accepted work stays at the gateway
Shell running hermes gateway run in the foreground The gateway depends on that process; do not rely on it surviving terminal closure or SSH logout
SSH connection to a server with an independently managed gateway service Disconnecting the client need not stop the service; check its actual status after reconnecting
The computer running Hermes Sleep or shutdown prevents that computer from executing work until it is available again

Closing a browser tab or a Desktop window is likewise not a useful health check. Find out which gateway/backend owns the work and whether it is still running. The upstream dashboard has its own scheduler support and defers to a live gateway for a profile; a saved job alone does not keep a backend alive.

For a self-hosted server, use a service and check it

Where you administer a supported Linux or macOS installation, upstream documents these commands:

hermes gateway install
hermes gateway start
hermes gateway status

On Linux, the default installation is a user service. For it to survive logout, the installation guide calls for an administrator to enable lingering for the actual service user: sudo loginctl enable-linger SERVICE_USER. Replace the placeholder with that user’s account name. Upstream also supports sudo hermes gateway install --system for a Linux boot-time system service. Use the installation instructions appropriate to your machine and service account.

After reconnecting, check the same user’s installation and profile. If you installed a system service, inspect it with hermes gateway status --system. A server that is powered on can still have a stopped or unhealthy gateway; an installed service can still need credentials or configuration repaired. For startup failures, use the VPS and Docker troubleshooting checklist.

Scheduled work needs more than a saved schedule

For a job you already have, start with reads:

hermes cron status
hermes cron list
hermes cron runs JOB_ID

Use the right profile and replace JOB_ID with the job’s ID. Check whether the scheduler is live, the job is enabled, when it last ran, and whether the expected output actually arrived. A successful run and successful delivery are separate observations.

hermes cron tick is one scheduler pass, not a command that leaves a daemon running. In this release, agent-backed jobs need an already-running gateway; a bare tick does not start one. Script-only no_agent jobs can run from a bare tick. An external trigger therefore does not remove every runtime dependency.

Upstream’s default local policy catches up a missed recurring job once when scheduling returns, rather than replaying every missed slot. Jobs do not run while paused; resuming a recurring job can trigger one catch-up run. cron.catch_up_missed: false changes the late-run policy. Do not assume a morning report ran at its original time because a result appears after restart. Inspect the run history and timestamp.

What remote hosting changes

If Hermes runs on a remote computer, closing your own laptop does not turn that computer off. It can still depend on a file, network connection or service available only at home; moving the agent does not move those dependencies automatically. The home-or-cloud guide helps with that decision.

Vavio supplies a managed Hermes computer for $14.99 per month, with model usage billed separately. That removes the need to maintain the hosting server yourself. It does not guarantee uninterrupted tasks, successful scheduled delivery or recovery within a fixed time. Check the current hosting offer, then judge your own job by its recorded run and delivered result. For existing owners, checking whether Hermes is working is the useful next step.

Sources

Written by Vavio with AI assistance and checked against pinned official Hermes v0.23.1 source and documentation. No terminal-close, logout or scheduled-job runtime trial was performed for this article.

More from the Vavio blog: Hermes Agent guides