A Hermes agent cron job lets a small business hand its repeat tasks, the morning check, the weekly report, the overnight watch, to an AI agent that runs them on a timetable without anyone pressing a button.
For most owners it's worth setting up, as long as you start with one job and set two safety rules from day one.
Those rules are new: you can now stop a job mid-run, and you decide what happens to runs your machine missed.
If you'd rather have my team work out which tasks to hand over first, book a free SEO strategy session.
The Short Answer For Business Owners
Yes, if you have a task that happens on a clock and follows the same steps each time.
No, if the task needs a judgement call you haven't written down yet.
Hermes Agent is open-source software from Nous Research, and cron is built into it.
You describe the task once, pick a schedule, and choose where the result lands.
The gateway, the always-on part of Hermes, checks for due jobs every 60 seconds, according to the official cron docs.
So the real question isn't whether it works.
It's which of your weekly chores you trust it with first.
What A Scheduled Job Does While You're Closed
Each run starts a fresh agent session, does the task, and delivers the final answer to a channel you pick.
That channel can be Telegram, WhatsApp, Slack, Discord, email, SMS or a local file.
You can set the job up in plain English.
The docs give the example of asking Hermes to check a news site every morning at 9am and send a summary to Telegram.
Hermes turns that sentence into a scheduled job through its own cronjob_manage tool.
No command line needed, though it's there if your tech person prefers it.
Use Case: A Local Shop's Opening Brief
Picture a shop owner who checks the same three things before opening: new online reviews, overnight orders and the day's deliveries.
That's a textbook first job.
hermes cron create "30 7 * * 1-6" \
"Check new reviews, overnight orders and today's deliveries. Reply in 5 bullets, problems first." \
--name opening-brief --deliver whatsapp30 7 * * 1-6 means 7:30am, Monday to Saturday.
The important bit is the prompt.
The docs are blunt that every run starts with no memory, so the prompt must contain everything the agent needs.
Add --continuity and each brief sees the last one, so it stops repeating yesterday's news.
Use Case: A Service Firm's Quiet Watchdog
A plumbing or IT firm doesn't want a report every hour.
It wants a message only when something breaks.
Hermes has a no-agent mode for exactly this: a small script runs on the schedule and its output is sent as the message.
Empty output means nothing is sent.
No AI model is called at all, so the check uses no tokens.
If the script itself fails, Hermes sends an error alert, so a broken watchdog can't stay silent.
For AI-driven checks there's a matching trick: tell the agent to reply with only [SILENT] when everything is fine, and that run delivers nothing.
Use Case: An Agency's Monday Client Roundup
An agency with clients in two countries can run one roundup per market.
The docs show per-job time zones, so a 9am job pinned to America/New_York fires at 9am there, even across clock changes.
You can also chain jobs.
With context_from, a "collect" job runs first and a "write the summary" job receives its latest output automatically.
The docs' own example is a three-stage chain: collect stories at 7:00, rank them at 7:30, draft posts at 8:00.
Swap stories for client metrics and you have a Monday pipeline your account managers read over coffee.
🔥 Want ready-made scheduled jobs for your kind of business? Inside the AI Profit Boardroom you get step-by-step video trainings, the installable Agent OS with Hermes inside, four coaching calls a week and 3,400+ members who share the jobs they schedule. → Join the AI Profit Boardroom
The Time It Actually Gives Back
I won't put a number on it, because it depends on the task.
What I can tell you is which steps disappear.
- Remembering to do the task at all.
- Opening the same four tabs every morning.
- Copying the result into a message for your team.
- Checking whether anything changed since yesterday.
The first one is the big one.
A task that happens on time, every time, beats a better task that happens when someone remembers.
How To Keep The AI Bill Predictable
Hermes has more spend controls than most owners expect.
- Bad setups don't run. Before any model call, Hermes checks the API key, skills and delivery targets. If something's wrong, the job is marked
blocked_config, you get one alert, and no tokens are spent. - Effort per job. A heavy weekly report can run at high reasoning effort while the hourly check runs at minimal.
- Skip when nothing changed. A small pre-check script can tell Hermes not to wake the agent at all for that run.
- Trim the tools. Giving a job only the web and file tools keeps every call smaller.
For unattended runs, the docs point to hermes setup --portal as the lowest-friction login, because it refreshes itself.
Keep Each Job On The AI Model You Chose
This one catches owners out.
A job that isn't pinned runs on whatever main model Hermes is set to at the moment it fires.
So if someone on your team switches the chat model on Tuesday, Wednesday's invoice check quietly runs on the new one.
The docs give you two fixes.
Pin a model to one job, or set a separate default for every scheduled job with cron.model, so chat changes never touch your schedule.
There's a trade-off worth knowing.
A pinned job won't switch to a backup provider if its own one is down, so the run fails and you get an alert.
An unpinned job can fall back to the backup list in your settings.
For a daily brief, I'd leave it unpinned so it still arrives.
For a report where the wording must stay consistent week to week, I'd pin it.
The New Stop Button For A Job That's Going Wrong
Here's the scenario every owner worries about: a job starts, and you realise the prompt is wrong.
Pausing doesn't help, because pause only blocks future runs.
You now have a proper stop.
In the Hermes Desktop app, a job that's running shows Stop in its header.
Press it, it shows Stopping…, and a Run cancelled toast confirms it.
Your tech person can do the same with hermes cron cancel <job> --reason "wrong prompt".
The stop-a-run section confirms the schedule stays as it was, the run is logged as cancelled with your reason, and it isn't counted as a failure.
What Stop Can't Undo, And The Rule That Covers It
Cancel stops the run.
It doesn't reverse anything the run already did.
If ten of twenty customer emails went out, those ten are out.
So for anything customer-facing, use an Ask first rule: the job drafts, a person approves.
Create the job paused with --paused, trigger one test run by hand, read the result, then switch it on.
The docs call this a safe canary, and it's the habit I'd insist on for any job that touches customers or invoices.
Missed Runs When The Office Computer Sleeps
If Hermes runs on a laptop that was closed at 7:30, the opening brief didn't fire.
By default it catches up once when the machine is back, and a long gap becomes one run, not a pile.
You can set that per job.
In the Desktop job editor, Missed runs offers three choices: Use the default, Run once when back, or Skip them.
| Job | Missed-run choice | Why |
|---|---|---|
| Opening brief | Skip them | A brief at 2pm is useless |
| Unpaid invoice check | Run once when back | Late is better than never |
| Weekly client roundup | Use the default | Catches up inside its window |
The default window is half the job's period, between 2 minutes and 2 hours, and --grace-minutes changes it, per the missed-run policy docs.
The cleaner fix is running Hermes on an always-on machine, so nothing is missed in the first place.
Holiday Mode For The Whole Business
Closing for a week? One command stops every scheduled job: hermes pause.
hermes resume lifts it, and work that came due catches up on the first tick.
The docs add a caveat: a run already going isn't killed, so cancel that separately.
Risks To Plan For Before You Rely On It
- The gateway is down. Jobs don't fire, and
hermes cron statusmarks them OVERDUE once a run is more than 15 minutes late. - The message didn't arrive. Hermes records
delivery_failedseparately from a failed run, so you know the work was done but not delivered. - A job keeps failing. After three failed runs in a row by default, the alert adds a nudge to fix, pause or remove it.
- The disk is full. Hermes skips runs it can't record and warns you, and
hermes doctorflags low disk space.
Before you start, update Hermes: the automatic retries after 5, 15 and 30 minutes for a run that couldn't reach the AI model arrived in v0.22.0, and the releases page lists everything since.
📞 Not sure which tasks to schedule first? Book a free SEO strategy session and my team will look at your week and point out the jobs worth automating. → Get your free strategy session
Who Sets It Up, Then A Sensible First Month
If you're comfortable typing a sentence into a chat window, you can create the job yourself.
Hand the gateway install and the always-on machine to whoever looks after your tech.
Then own the prompts.
You know what a good opening brief looks like; your tech person doesn't.
Here's the order I'd roll it out in.
- Week 1: one internal brief, created paused, tested by hand, then switched on.
- Week 2: add a quiet watchdog that only speaks when something breaks.
- Week 3: set the missed-run choice on both jobs.
- Week 4: add one weekly report at a higher effort level.
Leave anything customer-facing until the first three jobs have run cleanly for a while.
🔥 Get the Agent OS and the training to run it Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step videos on scheduling Hermes jobs for a small business, four coaching calls a week and 3,400+ members. → Join the AI Profit Boardroom
Also On Our Network
- 🌐 The seven-step operator setup, from gateway install to stopping a stuck run
- 🌐 An eight-row scorecard of Hermes scheduling and the final 8.3 verdict
- 🌐 Five prompts that make your agent schedule, stop and repair its own jobs
- 🌐 Which AI model to pin to each scheduled job, by real GoldieBench scores
- 🌐 The complete Hermes Agent Cron Job guide on the AI Profit Boardroom blog
FAQ: Hermes Agent Cron Job For Small Business
Do I need to code to create a Hermes cron job?
No. You can describe the task and schedule to Hermes in plain English and it creates the job itself. Someone still needs to keep the gateway running on an always-on machine.
What happens if a job sends the wrong message?
Stop it with the Desktop Stop button or hermes cron cancel. The run stops, but anything already sent stays sent, so customer-facing jobs should draft for approval first.
Can a job run without using any AI tokens?
Yes. No-agent mode runs a script on schedule and sends its output directly, with no model call at all.
What if my office computer was off when a job was due?
By default the job runs once when the machine is back. In the Desktop editor you can set each job to run once when back, skip missed runs, or use the default.
How do I stop every job for a holiday?
Run hermes pause to stop all scheduled fires and hermes resume when you're back. Runs already in progress need a separate cancel.
📺 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 a team of 70+ SEO experts.
- I've grown a YouTube channel to 400,000+ subscribers.
- I wrote "Link Building Mastery".
- My Udemy courses have taught over 29,000 students.
→ Get my best AI training inside the AI Profit Boardroom
Pick one repeat task this week, and if you want my team to help you choose it, book a free strategy session before you switch on your first Hermes agent cron job.
