Hermes Agent: The Free Jarvis Everyone Is Copying

Everyone wants their own Jarvis, the assistant from Iron Man that runs the house, answers back and handles things before Tony asks. Scroll YouTube for a minute and you'll find people building one. Many of the versions I see people building run on the same free piece of software, and it's called Hermes Agent.

Hermes Agent is an open-source AI assistant from Nous Research. You text it on Telegram, Slack, WhatsApp or a long list of other apps, it does the work, and it keeps running while you're asleep. The code is free under the MIT licence, and more than 255,000 people have starred it on GitHub.

I've run mine for almost half a year. I called it Steve. Here's what Hermes actually is, what I use it for, what other people build with it, and why setting it up is the hard part.

What is Hermes Agent?

Hermes Agent is an AI assistant that lives on a computer you leave switched on, instead of in a browser tab you open. That one difference changes what it can do.

A chat AI like ChatGPT waits for you. You open it, you type, it answers, and when you close the tab it stops. Hermes runs in the background on a server, so it can do things on a schedule, watch for something to happen and message you when it does, and pick up a job you texted it from your phone while you're out.

Its own docs list three things that make it work like that.

It talks to you where you already are. Hermes connects to Telegram, Discord, Slack, WhatsApp, Signal, SMS, email, Microsoft Teams, LINE and more than a dozen other messaging apps through what it calls a messaging gateway. You don't install a new app. You message it like a person.

It works on a schedule. You can tell it, in plain English, to do something every morning at 7am, every weekday at 9am or every Monday. Hermes turns that sentence into a scheduled job, and the result lands in your chat.

It asks before it does anything risky. On a messaging app, Hermes sends you the details of a dangerous action and waits for a yes before running it. If you don't answer within five minutes, the answer is no. That approval step is one of the controls Singapore's cyber agency asks for in its checklist for using AI agents safely.

What does a 7am summary actually look like?

This is the example I'd start with if you run a business. You tell it: every morning at 7am, send me a full summary of what happened overnight. From then on, it does that every day.

Instead of opening your inbox, your sales dashboard, your ad account and your messages one by one, which can easily be ten apps before breakfast, you read one message. Hermes's own docs use a version of this: "Every morning at 9am, check Hacker News for AI news and send me a summary on Telegram." Swap Hacker News for your own sources and the idea's the same.

What do I use mine for?

Steve handles a lot of the work for me now. It reads and sends my email, handles everything to do with finance, manages my social media, tracks client work and reporting, and watches the conversions on my website the way an SEO specialist would. And anything risky comes to me first in the chat for a yes.

The useful part is that all of it comes to one chat on my phone, and the boring checks happen without me having to remember them. I listed all ten jobs it does for me, and ten it could do for you separately. I also wrote about the memory side of it, how it knows my business without me re-explaining it every time, in the second brain behind my AI agent.

What are other people building with it?

The Jarvis trend on YouTube shows how many people want this. Damian Malliaros built a full Iron Man style voice interface on top of Hermes and walked through it in Turn Your Hermes Agent into JARVIS From Ironman. Plenty of other creators are building the same idea on other tools, so plenty of people clearly want an assistant that does things instead of just chatting.

Beyond the Jarvis look, here's the kind of thing Hermes is built to do, all from its own docs:

• Morning briefings and weekly reports delivered to whichever chat app you use.
• Watching for a problem and pinging you. The docs' example: "Ping me on Telegram if RAM is over 85%, every 5 minutes." Hermes writes the check itself.
• Connecting to your other tools. It supports MCP, the standard way AI assistants plug into apps, so it can reach things like Google Analytics, Stripe, HubSpot and Notion through their own connectors.
• Splitting a big job up. It can hand parts of a task to smaller helper agents that work in parallel.
• Learning as it goes. After a complex task it can save what worked as a reusable skill for next time.

If you came over from OpenClaw, Hermes's setup offers to import your old settings on the first run. I covered why so many people left OpenClaw in what happened to OpenClaw.

Is Hermes Agent really free?

The software is. The MIT licence means you can download it, run it and change it without paying Nous Research anything. Two things around it still cost money.

The AI model. Hermes is the assistant, not the brain. You connect it to a model provider like OpenRouter, OpenAI or Nous Research's own portal, and you pay that provider for what you use.

Somewhere to run it. For it to work while you're asleep, it has to live on a computer that stays on. The README says it can run on "a $5 VPS", a small rented server, or on services that sleep when idle and wake on demand.

So the software costs nothing, but running it does cost something.

Why is the setup the hard part?

If you're not a developer, setting Hermes up properly is a real project that takes far more than one click.

The install itself is one line in a terminal, curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash, and then a setup wizard. That part's fine. The hard part is everything after it: renting and securing a server so it can run all day, creating a Telegram bot and locking it so only you can talk to it, choosing a model and keeping the cost in check, connecting your email and tools one at a time, and then deciding what it's allowed to do on its own and what needs your yes.

None of that is in a single guide, because every business connects different tools. That's why I'd expect a lot of people to stop after the first chat. It's also the work I do for businesses that want one running without spending their weekends on it.

You tell me the jobs you want off your plate, I connect the tools and set what it can do alone, and the assistant shows up in a chat app you already use. Book a free 30-minute call and we'll work out which jobs it should take first.

FAQ

What is Hermes Agent?

Hermes Agent is a free, open-source AI assistant made by Nous Research. It runs on a server, talks to you through apps like Telegram, Slack and WhatsApp, runs jobs on a schedule, and asks for approval before risky actions.

Is Hermes Agent free to use?

The Hermes Agent software is free under the MIT licence. You still pay for the AI model it uses, through a provider such as OpenRouter or OpenAI, and for a server to keep it running, which its README says can be a $5 VPS.

Can Hermes Agent send me a summary every morning?

Yes. Hermes Agent has a built-in scheduler, so you can ask it in plain English to do something every morning at 7am, such as summarising what happened overnight, and it sends the result to your chat app each day.

Do I need to know how to code to set up Hermes Agent?

You don't need to write code, but you do need to be comfortable in a terminal. Installing Hermes Agent is one command, while running it all day on a server, securing the chat bot and connecting your tools takes real setup time.

Which apps can I talk to Hermes Agent from?

Hermes Agent's messaging gateway supports Telegram, Discord, Slack, WhatsApp, Signal, SMS, email, Microsoft Teams, LINE, Matrix and more than a dozen others.