agentos.guide › blog
Hermes Agent Background: The 10-Minute Quick-Start And How It Fits An Agent OS
Hermes Agent background work is anything the agent does while you aren't typing to it, and Hermes has seven built-in ways to do it.
The seven are cron jobs, the gateway service, /bg sessions, background terminal processes, delegated subagents, kanban workers and /loop.
Hermes Agent is free and open source from Nous Research, so the only running cost is the model you connect and the machine that stays awake.
Below you'll get a 10-minute quick-start, my real schedule, and how each background mode fits into an Agent OS.
Hermes Agent background work in one table
There are seven real ways to run Hermes in the background, and they aren't interchangeable.
| Way | Real command | What it's for | Survives a closed terminal |
|---|---|---|---|
| Cron jobs | hermes cron create | Work on a schedule with no chat open | Yes, while the gateway runs |
| Gateway service | hermes gateway install | Keeping Hermes itself alive | Yes |
| Background sessions | /bg <prompt> | A one-off task beside your chat | Only on a gateway chat |
| Background processes | Ask in plain words | Long shell commands the agent watches | No |
| Subagents | delegate_task, automatic | Parallel pieces of one big task | No |
| Kanban workers | hermes kanban create | Durable multi-agent tasks | Yes, the board is a database |
| Loops | /loop 5m <prompt> | Re-checking something on a timer | Only on a gateway chat |
Cron and the gateway are the two that matter most, because everything unattended sits on top of them.
The other five run beside a chat, and they're faster to start but less durable.
I checked each command against the official docs and my own install's help output, which is Hermes Agent v0.21.4, on 12 October 2026.
The 10-minute quick-start
This is the shortest path from a fresh install to a job that runs while you sleep.
Step 1: Install the gateway as a service
hermes gateway install
hermes gateway start
hermes gateway status
The gateway is the long-running Hermes process that answers chat apps, runs the scheduler and dispatches kanban workers.
The command installs a launchd agent on macOS, a systemd user service on Linux and a Scheduled Task on Windows.
On a Linux server you should also run sudo loginctl enable-linger $USER, so the service survives a logout.
The docs describe sudo hermes gateway install --system for headless hosts that should come back at boot.
Step 2: Create your first scheduled job
hermes cron create "0 6 * * *" "Read my to-do file and write a short recap of today's priorities." --name "Daily 6AM To-Do Recap"
You can also type /cron add "every 2h" "Check server status" in a chat, or ask Hermes in plain English.
The schedule accepts cron expressions, intervals like every 2h, delays like in 30m and natural forms like weekdays at 9am.
The docs say cron execution is handled by the gateway daemon, which ticks the scheduler every 60 seconds.
Each run starts a fresh session, so the prompt has to contain everything the agent needs.
Step 3: Run it once and read the result
hermes cron list
hermes cron run "Daily 6AM To-Do Recap"
The run command triggers the job on the next scheduler tick, and it accepts a job name or an ID.
Output from every run is saved under ~/.hermes/cron/output/, even if nothing is delivered.
Add --deliver telegram when you want the result sent to your phone.
Step 4: Know how to stop it
hermes cron pause "Daily 6AM To-Do Recap"
hermes pause
The first command pauses one job.
The second command is the global emergency stop, and it halts new cron dispatch, kanban dispatch and new gateway turns until you run hermes resume.
The help text says it never kills work that is already in flight.
🔥 Want the exact Hermes setup I run inside my Agent OS?
Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.
How Hermes Agent background work fits an Agent OS
An Agent OS is a dashboard that runs a crew of agents under one shared memory.
Background work is what makes that crew useful when you aren't at the desk.
I think about it as five layers.
| Agent OS layer | Hermes feature behind it | What I run there |
|---|---|---|
| The always-on layer | The gateway service | Chat apps, the scheduler and the board dispatcher |
| The schedule layer | Cron jobs | The 6am to-do recap, inbox jobs and the nightly notes summary |
| The parallel layer | Subagents and kanban workers | Research splits and multi-step builds |
| The watch layer | /loop and /heartbeat | Deploys, rankings and long jobs |
| The side-task layer | /bg and background processes | One-off research and long commands |
The always-on layer is the foundation, because cron and kanban both run inside the gateway.
The schedule layer is where the daily value comes from.
The other three layers are tools I reach for during a working session.
My real schedule
I ran hermes cron status on my Mac on 12 October 2026, and it reported the gateway running with five active jobs.
| Time | Job | What it does |
|---|---|---|
| 06:00 | Daily 6AM To-Do Recap | It reads my to-dos and writes the day's priorities. |
| 07:00 | Morning inbox summary | It summarises what arrived overnight. |
| 08:00 | Morning inbox triage | It sorts the inbox into what needs me and what can wait. |
| 18:00 | Evening wrap-up | It closes out the day's inbox work. |
| 21:00 | Nightly notes summary | It writes a summary into my notes vault. |
The list showed the 6am recap finishing at 06:01 that morning with the status ok and the dispatch marked on time.
My jobs are set to deliver locally, which saves each result as a file instead of sending a message.
In my Herald OS video I also showed a daily Japanese learning job in the same automations list.
One schedule, three front ends
Herald OS, Hermes Desktop and my Agent OS all sit on the same Hermes underneath.
In my Herald OS video the automations synced across all three, so the 6am recap showed up in each one.
That means you can create a job in the terminal and switch it off from a dashboard.
Herald OS also has a setting for the number of background agents and a setting for where agents run their tasks.
My Herald OS quick-start covers those screens.
The radar that wakes at 6am
My trend radar is the background job I'd miss most.
A scheduled job on my Mac asks Hermes to read the live X feed through Grok at 6am.
It returns six ranked AI stories with my content angle attached, pushes the top three to Telegram and logs the sweep to my notes vault.
The full build is in my Hermes Oracle Radar guide.
The other five background modes, quickly
/bg background sessions
/bg Analyse the logs in /var/log and summarise any errors from today
The /bg command runs one prompt in a separate session while your chat stays free.
The background session knows nothing about your current conversation, so the prompt has to be complete.
The result appears as a panel in the terminal, or it is sent back to the same chat on a messaging platform.
Background terminal processes
You ask for these in plain words, such as "run the build in the background and tell me when it finishes".
The agent's terminal tool starts the command, and its process tool can list, poll, wait for, read and kill it.
The /stop command kills all running background processes.
A process belongs to the agent that started it, so a subagent's processes end when the subagent finishes.
Delegated subagents
Hermes uses its delegate_task tool on its own when a task is complex enough.
Each child gets a fresh conversation and its own terminal, and only its final summary returns to the parent.
The docs say top-level delegations run in the background automatically, so the conversation continues while children work.
The defaults are 10 parallel children per batch and 250 iterations per child.
A Hermes restart does not resume a running child, so the docs point to cron for work that must survive one.
Kanban workers
hermes kanban create "Draft the weekly SEO report" --assignee researcher --max-runtime 30m
Every task is a row in ~/.hermes/kanban.db, and every worker is a full operating system process with its own profile.
The dispatcher runs inside the gateway, so step 1 of the quick-start already switched it on.
A task is blocked after two consecutive failed attempts by default.
/loop, /heartbeat and /goal
/loop 5m check the deploy status and tell me if it's live yet
The /loop command re-runs a prompt on a timer inside your current session, and each wakeup reads the state fresh.
If you leave out the interval, the loop paces itself from one minute up to a 15 minute ceiling while nothing changes.
/heartbeat every 10m <prompt> keeps one recurring instruction in a session, and /goal <text> keeps working until a judge model says the objective is done.
The docs say unattended overnight work should use cron, because these three live inside a session.
My Hermes /loop guide shows real sessions, and you may need hermes update on an older version.
How to see what's running in the background
| Question | Command |
|---|---|
| Is the scheduler alive? | hermes cron status |
| What is scheduled, and did it run? | hermes cron list |
| What happened on past runs? | hermes cron runs --limit 20 |
| Is any job unhealthy? | hermes cron doctor |
| Is the gateway up? | hermes gateway status |
| What did the agent just do? | hermes logs --since 1h |
| What did it cost? | hermes insights --days 7 |
| What is on the board? | hermes kanban list |
| Which subagents are running? | /agents in the TUI |
The TUI also has a live work monitor on Ctrl+T, which shows subagents and background processes together.
Herald OS adds an activity history screen, which is handy for spotting background work you forgot about.
If hermes cron status says a job is overdue, the scheduler has stopped ticking and hermes gateway restart is the fix.
🔥 Want the exact Hermes setup I run inside my Agent OS?
Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.
Approvals and safety for unattended runs
Nobody is there to click approve at 6am, so Hermes decides in advance.
| Setting | Default | What it means for an unattended run |
|---|---|---|
approvals.cron_mode | deny | A cron job that hits a dangerous command is blocked. |
approvals.unattended_mode | deny | Webhook and API sessions are blocked the same way. |
approvals.timeout | 300 | An unanswered approval is denied after 300 seconds. |
| Hardline blocklist | Always on | A fixed list of destructive commands can't be overridden. |
| Cron toolset | Set in hermes tools | Cron only gets the tools you allow for the cron platform. |
Hermes also scans cron prompts for prompt-injection and secret-exfiltration patterns when a job is created or edited.
A cron run can't create more cron jobs unless you switch on cron.allow_agent_scheduling.
Subagents can't ask you questions, write shared memory, send messages or schedule jobs.
Herald OS has an auto-approve switch with an "ask before sending messages" option, and I'd leave the asking on.
My own habit is to create a new job with --paused, run it once by hand and then resume it.
Cost control for background agents
Every scheduled run and every loop tick is a real agent turn, so frequency and model choice decide the bill.
| Lever | Effect |
|---|---|
cron.model | Every unpinned cron job runs on this model instead of your chat model. |
--model on a job | One job is pinned to its own model. |
--reasoning-effort minimal | A simple job thinks less and costs less. |
--no-agent --script | A script runs on the schedule with no model call. |
--monitor-url | The agent run is skipped when the page hasn't changed. |
delegation.model | Subagents run on an inexpensive model. |
loops.max_ticks | A loop pauses after 100 wakeups by default. |
goals.max_turns | A goal pauses after 20 continuation turns by default. |
The scheduler validates a job's setup before each run, and a misconfigured job is blocked without making a model call.
hermes insights --days 7 shows token usage and cost, and hermes usage shows your provider's rate-limit windows.
Herald OS has a Usage panel with tokens used, your plan and your token balance.
My /loop guide works through the backoff maths, where a fixed two-minute loop fires 240 times overnight and a self-paced one fires about 35 times.
Mac, VPS or Hermes Cloud for background jobs
A background job only runs while its machine is awake.
| Host | How it stays on | The catch |
|---|---|---|
| Your Mac | launchd runs the gateway and caffeinate blocks sleep. | Closing a MacBook lid still sleeps it. |
| A VPS | systemd runs the gateway at boot. | You own updates and security. |
| Hermes Cloud | Nous Research hosts the agent. | It bills by the day while running. |
The Hermes docs give this command to keep a Mac awake for as long as the gateway runs.
caffeinate -i -w $(cat ~/.hermes/gateway.pid) &
If the gateway was down when a recurring job was due, the job catches up once when the scheduler is back.
Docker is another supported route, and the official image restarts the gateway automatically if it crashes.
You can run hermes gateway run in tmux or under nohup for a quick test, but nothing restarts it after a reboot.
Hermes Cloud's Standard size cost $0.51 a day while running when I checked on 10 October 2026, with inference billed separately.
My Hermes Cloud guide covers the hosted route, and my best Hermes Agent VPS ranking covers servers.
What to read next
My Hermes Agent bot guide covers the chat apps that deliver these jobs to your phone.
My Hermes Agent backend guide covers where the agent's commands actually execute.
My best Hermes Agent use cases list has more jobs worth scheduling.
The official cron docs, delegation docs and kanban docs are the source for everything above.
Also On Our Network
FAQ
How do I run Hermes Agent in the background?
Install the gateway as a service with hermes gateway install, then create a job with hermes cron create. The gateway ticks the scheduler every 60 seconds, so jobs fire with no terminal open.
Does Hermes cron need the gateway?
Yes. The official docs say cron execution is handled by the gateway daemon. hermes cron status tells you whether the scheduler is alive and flags any overdue job.
What is the difference between /loop and cron in Hermes?
/loop re-runs a prompt on a timer inside your current session, which suits monitoring while you work. Cron runs each job in its own fresh session outside any chat, which suits unattended work.
How do I stop every background job at once?
Run hermes pause. It halts new cron dispatch, kanban dispatch and new gateway turns until hermes resume, and it never kills work already in flight.
Can I run Hermes Agent background jobs on a Mac?
Yes. The gateway runs under launchd, and the Hermes docs recommend the built-in caffeinate command to block sleep. Closing a MacBook lid still puts it to sleep, so use a VPS or Hermes Cloud for jobs that must run while the Mac is shut.
Final word on running Hermes Agent in the background
Install the gateway, schedule one small job and read its output tomorrow morning.
Once that first job has run on its own, you'll see why I build my whole Agent OS on top of Hermes Agent background work.
About Julian
I'm Julian Goldie, an AI entrepreneur, SEO expert and the founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
My YouTube channel has 400,000+ subscribers, and I run Goldie Agency, a seven-figure SEO agency.
Related guides on agentos.guide
→ Hermes /loop: The Command That Watches Your Business For You
→ The Hermes Oracle Radar: The Trend Machine That Wakes At 6AM
→ Best Hermes Agent VPS: My Pick + Server Quick-Start
Get the whole Agent OS
The installable system behind every guide on this site — dashboard, agents, memory and pipelines — updated as new tools land, with 3,400+ members and four coaching calls a week.
Join the AI Profit Boardroom →📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉