AI Income Desk Join AIPB

Hermes Agent Background Jobs: 7 Real Business Uses

Hermes Agent background jobs run on a schedule while you work. See 7 real business uses, the exact command for each, what they cost and how to stop them.

Julian Goldie — founder, AI Profit Boardroom
By Julian Goldie /16 min read
Get the AI Profit Stack Join AIPB →
  • 1,000+ done-for-you AI agent workflows
  • 5 live coaching calls / week with Julian
  • 7-day refund
  • 3,400+ AI operators inside

Hermes Agent background jobs are tasks the agent runs on its own, on a schedule or beside your chat, so the work gets done while you're busy with something else.

Hermes Agent is the free, open-source AI agent from Nous Research, and its background features are built in, so the only running cost is the model you connect and the machine it runs on.

A business owner cares about one question here, which is what work can leave your plate without you watching it.

This post answers that with seven real uses, the exact command behind each one, the safety settings for unattended runs and an honest look at cost.

I checked each command against the official Hermes docs and my own install's help output on 12 October 2026, and I make no income claims because I can't promise you a number.

What Hermes Agent Background Work Is, In Plain English

Think of Hermes as a member of staff who can be given standing orders.

A standing order is a job with a time attached, such as "every morning at six, read my to-do list and send me the three things that matter".

Hermes calls that a cron job, and it's the main background feature.

There are six other background features, and each one fits a different kind of work.

Background feature What it is in one line The business job it fits
Cron jobs A prompt that runs on a schedule in a fresh session Daily and weekly recurring work
Gateway service The Hermes process that stays on after you close the terminal Being reachable on Telegram or Slack all day
/bg sessions A separate session that works beside your chat A one-off research task while you carry on
Background processes Long shell commands the agent starts and watches Builds, exports and scripts that take minutes
Subagents Child agents that take pieces of a big task Parallel research on several competitors
Kanban workers Tasks on a durable board, each run by its own agent process Multi-step work with handoffs and retries
/loop A prompt that re-runs on a timer inside a chat Watching something until it changes

The one thing every durable job depends on is the gateway.

The official docs say cron execution is handled by the gateway daemon, which checks the schedule every 60 seconds.

You switch it on once.

hermes gateway install
hermes gateway start

After that, your jobs fire whether or not a chat window is open.

🔥 Want my exact Hermes Agent Background Jobs setup, step by step? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom

Use 1: A Daily 6am To-Do Recap

This is my own first job of the day, and it's the one I'd tell any owner to copy first.

My job is named Daily 6AM To-Do Recap and it uses the schedule 0 6 * * *.

When I ran hermes cron list on 12 October 2026, it showed that this morning's run finished at 06:01 with the status ok.

A command for a job like mine looks like this.

hermes cron create "0 6 * * *" "Read my to-do file and write a short recap of today's priorities, deadlines and anything overdue." --name "Daily 6AM To-Do Recap" --deliver telegram

The value for a business is simple, because you start the day with a plan that you didn't have to build.

The prompt has to be complete in itself, because a cron job starts a fresh session with no memory of your earlier chats.

Use 2: Inbox Triage Before You Sit Down

My install has two inbox jobs, a 7am inbox summary and an 8am inbox triage.

Both ran on time this morning according to the same hermes cron list output.

The idea is that the agent reads what arrived overnight and sorts it into what needs you, what can wait and what is noise.

I'd keep one rule for this job in any business.

The agent should draft and sort, and a human should send.

That keeps a background job from replying to a customer in a way you wouldn't.

Hermes helps here, because dangerous commands are denied by default in cron runs through a setting called approvals.cron_mode.

A job that needs a tool it shouldn't have can also be limited with hermes tools, where cron has its own tool list.

Use 3: A Trend Radar That Wakes Before You Do

Content businesses live on timing, so I built a radar.

A scheduled job on my Mac asks Hermes to read the live X feed through Grok at 6am.

It comes back with six ranked AI stories, each with a reason it is trending and my content angle attached.

The top three are pushed to Telegram, and every sweep is logged to my notes vault.

I documented the whole machine in my Hermes Oracle Radar guide.

The business lesson isn't the tool, because the lesson is that research with a fixed shape is perfect background work.

If you can describe the report you want every morning, an agent can have it waiting.

Use 4: A Daily Practice Drill

Not every background job is about the business.

In my Herald OS video I showed a daily Japanese learning job sitting in the same automations list as my to-do recap.

I'm learning Japanese, so a short drill that arrives every day keeps me consistent.

The same pattern works for staff training, such as one product question a day sent to a team channel.

A job like that looks like this.

hermes cron create "every day at 9am" "Write one short Japanese practice drill with five phrases, a translation and one question for me to answer." --name "Daily Japanese" --deliver telegram

Hermes accepts plain schedules like every day at 9am as well as cron expressions.

🔥 Want my real automations, prompts and the Agent OS they run in? Inside the AI Profit Boardroom you get my Agent OS, step-by-step Hermes training and four coaching calls a week with 3,400+ members. → Join the AI Profit Boardroom

Use 5: Watching A Deploy, A Ranking Or A Long Job

A lot of an owner's day is checking whether something has finished.

I run an SEO pipeline across five websites, and most of it is waiting for something to land.

The /loop command hands that waiting to the agent.

/loop 5m check the deploy status and tell me if it's live yet

Hermes re-checks the real state on each wakeup and tells you when it changes.

If you leave out the interval, the loop paces itself from one minute up to 15 minutes while nothing changes.

You can stop it with a condition, such as --until queue depth reaches zero, or a cap, such as --times 30.

The docs say /loop lives inside a session, so overnight work still belongs in cron.

My Hermes /loop guide shows real sessions.

Use 6: A Weekly Report Built By Several Agents

Some work needs more than one pass, such as a researcher, a writer and a checker.

Hermes has a kanban board for that.

Every task is a row in a local database, and every worker is its own agent process tied to a profile.

hermes kanban create "Draft the weekly client report" --assignee researcher --max-runtime 30m
hermes kanban list

The board dispatcher runs inside the gateway, so step one of this post already switched it on.

A task that fails twice in a row is blocked instead of retried forever, which protects your budget.

For smaller splits, Hermes delegates to subagents on its own.

The docs say top-level delegations run in the background automatically, so your chat stays usable while the children work.

An agency could use that to research five competitors in parallel and get one summary back.

Use 7: A Watchdog That Costs Nothing To Run

Some checks don't need a model at all.

Hermes has a no-agent mode, where a script runs on a schedule and its output is delivered as the message.

hermes cron create "every 5m" --no-agent --script memory-watchdog.sh --deliver telegram --name "memory-watchdog"

If the script prints nothing, nothing is sent, so you only hear from it when something is wrong.

The docs state that this mode uses no tokens and no model.

A site-down check, a low-disk alert or a "the form stopped working" check fits this pattern.

🚀 Ready to put Hermes Agent Background Jobs to work in your business? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom

Which Background Mode Fits Which Job

The job The mode I'd use Why
A report at a fixed time Cron It runs with no chat open and survives restarts.
A one-off task while I keep chatting /bg It runs in a separate session and reports back.
Waiting for something to change /loop It re-checks fresh state on a timer.
A long export or build A background process The agent starts it and tells me when it ends.
Research across many sources Subagents The pieces run in parallel.
Work with handoffs and retries Kanban The board is durable and tracks every step.
A simple alert No-agent cron It spends no tokens.

How To Check On Background Work And Stop It

You wouldn't hire someone and never look at their work, and the same goes for an agent.

hermes cron status tells you whether the scheduler is alive and when the next job fires.

hermes cron list shows every job with its last run and status.

hermes cron runs --limit 20 shows the history of attempts.

hermes logs --since 1h shows what the agent did in the last hour.

hermes insights --days 7 shows token usage and cost for the week.

Herald OS, the free desktop shell I reviewed, also has an activity history screen that shows background work you forgot about.

Stopping is just as quick.

hermes cron pause <job> stops one job.

hermes pause is the emergency stop, and it halts new scheduled work, new board work and new chat turns until you run hermes resume.

The help text says it never kills work that is already in flight.

Safety: What An Unattended Agent Is Allowed To Do

The fear with background agents is that one does something silly at 3am.

Hermes handles that with defaults that lean towards no.

A cron job that reaches a dangerous command is denied by default, and the same applies to webhook and API sessions.

An approval request that nobody answers is denied after 300 seconds.

A fixed blocklist of destructive commands can't be bypassed by any setting.

Hermes also scans cron prompts for prompt-injection and secret-stealing patterns when a job is created.

A cron run can't create more cron jobs unless you switch that on yourself.

Herald OS has an auto-approve switch with an "ask before sending messages" option, and I'd keep the asking on for anything a customer could see.

My own extra rule is to create a new job with --paused, run it once by hand, read the output and only then resume it.

🔥 Want someone to check your automation plan before you switch it on? The AI Profit Boardroom has four live coaching calls a week, my full Agent OS and 3,400+ members building real automations. → Get access here

What Background Jobs Cost A Business

Hermes itself is free, so the bill has two lines.

The first line is the model, because every scheduled run is a real agent turn.

The second line is the machine that has to be awake.

Cost lever What it does
cron.model It runs every unpinned scheduled job on a cheaper model than your chat.
--model on one job It pins one job to the model that suits it.
--reasoning-effort minimal It lowers the thinking level for a simple job.
--no-agent It runs a script with no model call at all.
--monitor-url It skips the agent run when the page hasn't changed.
delegation.model It runs subagents on an inexpensive model while the parent plans on a strong one.
loops.max_ticks It pauses a loop after 100 wakeups by default.

The frequency matters more than people expect.

My /loop guide works through the maths, where a fixed two-minute loop fires 240 times overnight and a self-paced one fires about 35 times.

A job that runs once a day is a rounding error next to a job that runs every two minutes.

Hermes also checks a job's setup before each run, and a misconfigured job is blocked without making a model call.

Mac, VPS Or Hermes Cloud?

A background job only runs while its machine is awake.

I run mine on a Mac, because my Agent OS and my local models live there.

The Hermes docs recommend the built-in caffeinate command to keep a Mac awake, and they warn that closing a MacBook lid still puts it to sleep.

If the gateway was down when a recurring job was due, the job catches up once when it comes back.

A VPS is a rented server that never sleeps, and I ranked the options in my best Hermes Agent VPS post.

Hermes Cloud is the hosted version from Nous Research, and its Standard size cost $0.51 a day while running when I checked on 10 October 2026, with model usage billed on top.

Third-party hosts exist too, and I looked at one in my FlyHermes post.

My advice is to start on the computer you already own and move a job to a server only when it must run while you sleep.

Where To Go Next

If you want the agent to answer you on Telegram or Slack while these jobs run, read my Hermes Agent bot post.

If you want to know where the agent's commands actually execute, read my Hermes Agent backend post.

If you want more jobs worth handing over, my best Hermes Agent use cases ranking has them.

FAQ: Hermes Agent Background Jobs

What are Hermes Agent background jobs?

They are tasks Hermes runs without you typing to it.

The main kinds are cron jobs, the gateway service, /bg sessions, background processes, subagents, kanban workers and /loop.

What can a business use them for?

Good first jobs are a morning recap, inbox triage, a trend radar, a daily drill, a deploy watcher, a weekly report and a watchdog script.

How much do they cost?

Hermes is free, so you pay for the model each run uses and the machine that stays awake.

A script-only job uses no model at all.

Can Hermes send messages on its own in the background?

A cron job delivers its final response to the target you choose, such as Telegram, Slack or email.

I'd have it draft customer replies for review instead of sending them.

Do I need a server?

No, because the jobs run on your own computer while it's awake and the gateway is running.

A server or Hermes Cloud only matters when the jobs must run while your computer is off.

Also On Our Network

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉

About Julian

I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.

I help business owners scale with AI agents, automation and SEO.

  • I've built a 7-figure SEO and link-building agency, Goldie Agency, with 70+ people in Bangkok.
  • I've grown a YouTube channel to 400,000+ subscribers.
  • I wrote the book "Link Building Mastery", which covers nine link building strategies.
  • My Udemy courses have taught 29,000+ students.

→ Get my best AI training inside the AI Profit Boardroom

Pick one job from the list, schedule it tonight and check the output tomorrow, because that's the fastest way to see what a Hermes Agent background job can take off your plate.

Member wins · screenshots

Real wins from inside the AI Profit Boardroom

See all 3,400+ members →
AIPB member win screenshot 1 AIPB member win screenshot 2 AIPB member win screenshot 3 AIPB member win screenshot 4 AIPB member win screenshot 5 AIPB member win screenshot 6 AIPB member win screenshot 7 AIPB member win screenshot 8 AIPB member win screenshot 9 AIPB member win screenshot 10 AIPB member win screenshot 11 AIPB member win screenshot 12

Your next move

Ready to build AI agents that actually make money?

Join 3,400+ entrepreneurs inside the AI Profit Boardroom. Get 1,000+ plug-and-play AI agent workflows, 5 live coaching calls a week, and a community that holds you accountable.

Join the AI Profit Boardroom →

7-day no-questions refund · Cancel anytime

Found this useful? Get more of this site in your Google.

One tap adds us as a preferred source — Google then shows you our posts more often in Search, Top Stories, and AI Overviews.

Or add us in your Google settings →

← Back to all posts