Hermes Agent: How to Set Up Your Private AI Assistant

By Thomas Echezabal, AI Automation Coach, thomasknows.ai

Key Takeaways

In this guide, you'll learn:

• What Hermes Agent is, how it differs from the Nous Hermes models, and why a private assistant with real memory matters
• How it works under the hood — one engine that holds memory, skills, and scheduled tasks, wrapped around any AI model you choose
• How to download and set up Hermes Desktop in a few minutes, from opening the installer to a working assistant you can chat with
• What you can do from inside the app's own Skills and Cron panes, entirely through clickable menus
• The skills, tools, and habits that separate a useful Hermes from an expensive toy

What Hermes Agent Is (and What It Is Not)

Hermes Agent is an open-source AI assistant from Nous Research that you download and run as Hermes Desktop, a native app for Mac, Windows, and Linux. You talk to it the way you'd text a capable employee, and it remembers what you told it last week, keeps skills it has picked up, and works from a chat window built specifically for it. The whole thing is free and released under a permissive open-source license, so the only thing you ever pay for is the AI model behind it.

One point trips people up immediately, so let me clear it up first. Nous Research also publishes a family of AI models called Hermes, and those are a separate thing entirely. The models are the raw brainpower; Hermes Agent is the assistant built around a brain, and Hermes Desktop is the app you use to talk to it. Hermes Agent can run on a Nous Hermes model if you want, but it works just as happily on Claude, GPT, Gemini, or other providers you connect. When I say "Hermes" in this guide, I mean the agent — the assistant you download and live with through its desktop app.

What makes it worth trying is what you rarely get in one tool: memory, because it carries your context forward instead of starting cold every time; choice, because nothing about it locks you into one company's model; and a real interface built for the job, with a chat window, a model picker, and panes for skills and scheduled tasks instead of a single input box. For a business owner who has grown tired of re-explaining themselves to a chatbot that forgets, that combination is the whole point.

What It Actually Does for You

Strip away the technical framing and Hermes earns its keep in a few concrete ways. The memory is what you notice first — a conversation on Friday already knows what you decided on Monday, so you stop re-explaining your business every time you open a chat. The app runs as its own program, not a browser tab, so that context stays put between sessions. What compounds over time is the learning: walk it through a task once, and it can keep that as a skill it reuses for good.

I can open Hermes Desktop and say, in plain language, "every weekday morning, check my priorities and send me a short brief." From there, the assistant keeps that as a standing task inside its own scheduling pane, running it the way I asked without me retyping the request. That shift — from a tool you operate one prompt at a time to an assistant that keeps working on your behalf — is what people mean when they call Hermes an agent rather than a chatbot.

How Hermes Works

Underneath the chat window, the design is simple enough to hold in your head. One assistant runs inside the Hermes Desktop app on your machine. It keeps your memory, holds the skills it has learned, tracks your scheduled jobs, and reaches out to whatever AI model you chose to do the actual thinking. You talk to it from the app's own chat interface, and it coordinates everything else behind the scenes.

The piece in the middle is the part that matters. Because the app holds your memory and your skills between sessions, it can do things a normal chat window cannot — pick up where you left off and improve at your work instead of resetting to zero every time you open it. The model on the right is swappable on purpose. You can start on a frontier model for the best quality, drop to a cheaper one for routine jobs, or connect a different provider entirely, all without losing anything the assistant has already learned.

What You Need to Run It

The requirements are lighter than most people expect, because Hermes itself does not do the heavy AI thinking — it hands that off to a model. Hermes Desktop runs on Mac, Windows 10 and 11, and Linux. The heavy AI thinking happens in the model you connect to rather than on your own machine, so an everyday computer is the expected setup.

You do need two things. The first is an everyday computer you can leave the app open on while you're using it. The second is an AI model for it to think with, which is the one part that usually costs money — either a subscription that bundles models together, or a pay-as-you-go account with a provider like Anthropic or OpenAI. You choose that inside the app itself, and you can change it anytime without starting over.

Setting Up Hermes Desktop

Getting a working assistant running takes a few minutes the first time, and every step happens inside the app itself.

Video: Hermes Agent setup walkthrough, from Alex Finn

Step 1: Confirm your device is supported

Make sure you're on a Mac, a Windows 10 or 11 machine, or Linux. That's the only check that matters before you download anything.

Step 2: Download and open Hermes Desktop

Go to the official Hermes Agent site and download the installer for your operating system, then open it the way you'd open any other app you've downloaded. On first launch, the app walks you through a short welcome screen.

Step 3: Connect a model

Hermes needs a brain, and you get to choose it from inside the app. The onboarding screen lets you sign in and connect a provider right there, or choose "provider later" if you'd rather explore the app first and connect a model when you're ready. Either way, this is the step where you decide how smart and how expensive your assistant is, and the app's Providers settings pane lets you change it anytime by signing in and storing credentials for a different provider.

Step 4: Start your first conversation

Once a model is connected, type into the chat window the way you'd text anyone. Hermes streams its answer back as it thinks, and shows you what it's doing along the way — if it's searching, reading a file, or using a tool, you see that activity appear in the conversation rather than waiting on a blank screen. You can drag a file straight into the chat to hand it something to work with, and a preview rail on the right side of the window shows web pages, files, and tool output without you leaving the conversation.

Step 5: Add your first skill

Open the Skills pane inside the app to browse the built-in catalog and install what matches your work. You'll find ready-made skills bundled with the app, note-taking among them. Start with one — the fastest way to feel the difference is to add the skill that touches the task you do most often, then ask Hermes to do that task and watch it use the skill on its own.

Step 6: Set up a recurring task

Open the Cron pane to view and manage scheduled jobs, like a morning brief or a recurring check-in, so recurring work runs on a schedule instead of waiting for you to remember it.

The Skills Worth Starting With

Hermes ships with a catalog of skills, and the temptation is to install a handful on day one. Resist it. The people who get the most out of Hermes start with the one or two skills that touch their daily work and let the rest come later. The safest bets are the skills that ship alongside the assistant itself, maintained as part of the same project, with the bundled note-taking skill a natural first pick.

The capability worth understanding early is how the skill system learns: hand Hermes a document, a link, or a walkthrough of a task, and it can turn that into a reusable skill on the spot, browsable and installable right from the Skills pane in the app. My advice is to get real value from one or two built-in skills first, learn how Hermes uses them inside a real task, and only then go looking for more. The goal is an assistant that does your actual work, not the biggest skill collection on the block.

Getting the Best Results

The owners I've watched try agents obsess over which model to use and ignore how they've organized the work, when the second one matters far more. The way I see it after running this myself: a powerful agent paired with a messy request is just an expensive intern. The setup is what carries you, not the horsepower.

A few habits separate a Hermes that earns its keep from one that frustrates you:

• Be specific about what you want remembered. Hermes keeps a deliberate memory rather than hoarding everything, and that's a feature. Tell it plainly what matters about you and your business, and it carries that forward.
• Match the model to the job. Use a strong frontier model for thinking-heavy work and a cheaper, faster one for routine chores, right from the model picker in the chat window. You're not locked in, so there's no reason to pay frontier prices for a task that doesn't need it.
• Let it work on a schedule. The single fastest way to feel the value is a recurring task set up through the Cron pane. Tell it plainly what you want and when, and let it handle the timing.
• Start narrow with skills. Add the one skill that touches your most frequent task before you go browsing the full catalog. A pile of half-used skills adds noise, not power.

Do those four things and model quality stops being the deciding factor. From running my own setup, I keep landing in the same place: clear instructions and good scheduling beat raw model power almost every time.

Common Pitfalls to Avoid

Hermes rewards patience, and the people I've watched bounce off it usually hit the same few snags. Knowing them in advance takes most of the sting out.

• Chasing the cheapest model. A bargain model that needs three tries to get something right can cost more than a good model that nails it once. Pick for reliability on the task, not for the lowest sticker price.
• Installing too much at once. A pile of half-used skills adds noise, not power. Add them one at a time as a real need shows up.
• Skipping the onboarding screen. Choosing "provider later" is fine if you want to look around first, but connect a real model before you expect much from it — an assistant with no brain attached can't do the parts that matter.
• Expecting it to feel identical to Claude Desktop on day one. Hermes asks a little more of you in exchange for choice of model and a memory that carries across sessions. Give it a real task before deciding whether that trade is worth it for you.

Hermes Agent vs Claude Desktop

People often ask how Hermes compares to a polished commercial assistant like Claude Desktop, and the honest answer is that they're built for different temperaments. Claude Desktop is effortless and answers the moment you ask, but it runs only on Claude's models and forgets each conversation when it ends. Hermes Desktop asks a little more of you up front and gives more choice and memory in return. If you've used Claude and want the broader picture of what it does well, my complete Claude guide for small businesses covers that side in depth.

The most useful way to see them is not as rivals. You can run Claude as the brain inside Hermes, pairing Claude's reasoning with Hermes's memory and skills. Many of the people getting the most out of this run both — a polished assistant for deliberate deep work, and Hermes for the tasks that should keep running on their own.

Frequently Asked Questions

These are the questions I hear most when someone is deciding whether Hermes is worth their time.

Is Hermes Agent free?

The software is free and open source under the MIT license, so you only pay for the AI model it uses. Budget models run a few dollars a month; frontier models like Claude or GPT cost more.

Is Hermes Agent the same as the Nous Hermes models?

No. They share a brand name but are different things. The Nous Hermes models are language models you can run, while Hermes Agent is the assistant that runs around a model, remembers your context, and does the work. Hermes Agent can use the Hermes models, but it can also use Claude, GPT, Gemini, or a local model.

Do I need to know how to code to use Hermes Desktop?

No. Hermes Desktop is a downloadable app with a chat window, a model picker, and management panes for skills and scheduled tasks. You download it, open it, and talk to it in plain language, the same way you'd text anyone.

Is my data private with Hermes?

Hermes runs on your own machine, and your data stays there by default. The exceptions are services you choose to connect, and the AI model provider you pick, which sees what you send it.

Can I run Hermes Desktop on a Mac?

Yes. Hermes Desktop runs on Mac, Windows 10 and 11, and Linux.

What can I do from inside the Hermes Desktop app?

You can chat with your assistant, switch models, drag files into the conversation, preview web pages and tool output side by side, browse and install skills, and view and manage scheduled tasks, all from the app's own windows and panes.

How is Hermes Desktop different from Claude Desktop?

Claude Desktop is a polished app that answers when you ask and runs only on Claude's models. Hermes Desktop lets you pick any model, keeps memory and skills that carry across sessions, and gives you management panes for automations that Claude Desktop does not offer.

Want This Set Up for You?

Hermes is one of the most capable private AI assistants you can run, and getting it tuned to your business takes a little more thought than the default setup. If you'd rather skip the trial and error and just get a working assistant tuned to your business — connected to the right model, loaded with the right skills, and handed to you ready to use — that's exactly the kind of setup I do. Book a 1:1 session with me and we'll map out exactly what your setup needs.

About the Author

Thomas Echezabal helps small business owners automate their busywork with AI and get real hours back each week. He has worked with small businesses his entire career, including 200+ clients on Fiverr.