Vollständiges Transkript anzeigen (2.876 Wörter)
Welcome to the Hermes Agent Masterclass.
This is Module 1. We're going to be starting literally from zero. And this is going to be the first part of a 10-part series that'll take you from zero to having multiple agents, multiple agentic workflows, and just about everything you would need to master Hermes Agent.
Now, I did this first video back in April and obviously a lot has changed since then. It's currently July 2026, so I wanted to do an update video. So a lot of this will remain the same, but some of it will be changed to reflect the changes in Hermes Agent since April. So by the end of this module, you'll have Hermes installed, configured with a working AI provider, you know the 20 commands you'll actually use everyday, and most importantly, you'll have completed a real task with it. Not a toy example, a real report you can actually use.
We're going to start out by explaining what is Hermes Agent. Uh, we're actually going to go through the installation and setup wizard. I'm going to explain CLI commands top to bottom and then we're going to have a demo, our first real task with the agent, and then a wrap-up and talking about what you'll see next in this whole masterclass. So let's get started.
And if you like this video, please consider following me on X at TonbiStudio. I'll often be posting short videos on the latest AI news and agent features. Also sign up for my free weekly newsletter, which I write by hand and release every Friday. Here, you'll see my honest thoughts about the latest AI news, models, research, and give a sneak peek of new projects I'm working on before they are announced anywhere else. Sign up on UnchainAIGarage.com, link is in the description.
So what is Hermes Agent? Hermes Agent is an open-source AI agent, it's built by Nous Research. They are the lab behind the Hermes model family. Most agents are built by tool companies, this one is built by people who train models, and that matters. And I'm going to talk about that a bit more in a second. It runs in your terminal, uh whereas a gateway it can answer through your phone, via Telegram, Discord, Slack, and a bunch of other messaging platforms. They also have a new desktop app that you can use to access it there.
It works with any model. Claude, GPT models, Grok, Kimi, Gemini, any local model. Hermes doesn't care. You can take your pick. At this point, it has over 200,000 stars, nearly 2,000 contributors total, and we're up to version 18. And they generally ship new versions every week or so. So here's what makes it architecturally different from others.
So the key bet of Hermes Agent is compounding value. Most agents are stateless. So every session starts from zero. Every ChatGPT session starts blank, every new Claude conversation starts blank, open Claude usually the same. Hermes was built differently. Every session feeds the next one. So there's a 4-step loop here.
You give it a task, it could be anything. It records a trajectory, and this is the tools used, every decision made, every result, and then it can extract that as a reusable skill based on how it worked. So the next time it's faster, more accurate and costs less tokens. And this is compounding value as a thesis and what makes Hermes really unique. And Hermes installed that you've had for 3 months is fundamentally better than what you set up yesterday.
So you're going to get to see this for yourself. I'm going to be installing this on a new device that has never had a Hermes agent and through this masterclass, you're going to be able to see exactly how it develops, what skills it learns, and how it becomes smarter.
So how do you know Hermes Agent is right for you? So if you use agents enough that compounding value pays off, occasional users may not feel it. If you want model freedom, you might want to start on Claude, want to start on Grok, end up using local, Hermes can roll with that pretty easily. And if privacy or cost matters to you, the local model support is really first class. So you may not want Hermes if you just want a chatbot, you could just use ChatGPT, you don't need a full agent.
Okay, so let's get into the installation here. You can go to NousResearch/hermes-agent on GitHub and get all of the information you need. Uh, down here you'll see the different ways to install it. It's available Linux, macOS, WSL, Termux, and uh recently, they have a Windows native version. And I'm actually going to install this in Windows, which is this one command in your PowerShell.
So I put this in here in a terminal, and it started up the Hermes Agent installer, you can see now it's going to install all of the uh dependencies that I need. And this should only take a couple minutes. Okay, so it's done, it's starting up a Hermes Agent setup wizard. I had an old OpenClaw installation on here, so you can migrate that, but I'll just say no for now.
So there's a couple different options, right, when you go to setup. You can do a quick setup, uh which is just going to create a account through new portal and you get a free login and this only takes like 30 seconds doing the quick setup. Blank slate here, this is everything off except the bare minimum. So this is really starting from a blank slate.
I would only do this if you're very proficient with agents and know exactly what you want. Uh, what we're going to do here is we're going to do 2, which is the full setup, so that I can kind of show you through the whole the whole setup. So let's see.
So the first thing it'll go through is the provider list, obviously you need a model to get an agent working. And new portal is the kind of default, um but you can see every other option in here. OpenRouter is a very popular one, you can always as well set up local models here. Uh, at module 5, we'll cover that. OpenAI, you can use your Codex Oauth or Grok Oauth here, and there's a bunch of different options that you can see.
So I do have a new portal account, so I'm just going to go with 1. Um, this will pop up this in the browser, right, your subscription. You can create a free account just with Google login or whatever. And this will pop up. You can just connect here. Uh, I have a plus account, but there's a free tier as well. And there we go. Login successful. So then you choose your model, and this can be easily changed afterwards. Um let's go with the uh 26, which is GLM 5.2. Pretty good model.
So now your new subscription includes tool gateway. You can choose which tools to enable. Um there's all of these different options, and you can configure all of these, for example image generation. I think the default here is FAL, but you can change this to uh GPT images or Grok Imagine, and you can adjust all of these as is, but I'll go more into that into the tools module.
So terminal backend, most of us are going to be doing local, that just means running directly on your machine. Um you could also do it with a Docker backend, that is going to be in module 2, I'll talk about the different backends, but we are just going to keep it uh the default. Messaging platforms is another one that we'll be talking about in module 2, uh but you can see all of the different um messaging platforms that you can use. For now, uh we just will leave them unconfigured.
Tool configuration for the CLI is here. So you can see the default ones that are allowed. You can choose to remove or add some of these. For example, if you have um the XAI Oauth or an XAI API key, you can use X or Twitter search, uh we don't have that set up right now, so we won't do that. Uh, I would generally leave this mostly blank uh at the default for now.
So browser automation is a new one. Uh, let's do local browser. It's recommended and free. And then computer use is installing. So next up is text-to-speech, and uh new subscription, this comes with the new portal, um so you can use that one. There's also Microsoft Edge TTS, which is free. Uh, pretty good quality. And there's a bunch of different options. I know 11 Labs is very popular, XAI has a really good uh TTS. There's a bunch of different options, just keep it in the defaults. And next we have web search and extract, select search provider.
Web search, very popular ones. Um, it comes, there's one package in the new subscription itself. There's other free options as well, Firecrawl self-hosted. I know DuckDuckGo is very popular, Brave search, uh Tavily as well is very popular. I've used this one a lot. Um it says paid but you can get an API key, there's a free tier, and the limits are pretty generous so you can do that as well.
And that's it. So to get started, to get talking to your agent you just do Hermes chat, pretty simple, and this will pop up uh the chat in the terminal, which is also called TUI. If you hear me referring to TUI in these masterclass videos, that's what I mean. And you can see the very iconic Hermes Agent kind of icon here. And this will show you uh what your model is, um as well as your available tools, available skills, and everything like that. Your context window.
So that was the TUI, and just to very quickly show you a couple nice features that have been added in the months since I did the first video. Hermes Dashboard is going to bring you to this web dashboard that should pop up on your browser, and there you go. The first time you do it, it'll take a little bit of time, but it popped up right in my browser, um you can see different sessions, uh files, models, cron jobs, this has basically everything you would need for the config, um if you want to change things in here.
And it's just a really great way, just in your browser to see everything that you have. Um so, you don't have to use the the terminal itself to change anything in the config, you can just do it through the web dashboard. And I mentioned this earlier, but if you do Hermes Desktop in the terminal, it'll automatically install the desktop uh desktop app. This will pop up.
And if you want to know more about the desktop app, check out my complete guide to the Hermes Desktop App. I have a lot of information, but this will pop up and it's a nice little app, um if you've ever used Codex, something similar like that. You have all your sessions here, and you can change the configurations as well in the settings, uh it's really convenient. And I mentioned this earlier, but if you do Hermes Desktop in the terminal, it'll automatically install the desktop app.
And that's it. So to get started, to get talking to your agent you just do Hermes chat, pretty simple, and this will pop up uh the chat in the terminal, which is also called TUI. So that was the TUI, and just to very quickly show you a couple nice features that have been added in the months since I did the first video. So, you can browser, browse which session you want to resume. Hermes Skill Browse, you can discover and install skills, let's do that.
Hermes config show, which is going to show your configuration, like I said Hermes Doctor and Hermes Update, which pulls the latest. Let me see if we're actually updated yet. Hermes Update. I think they may have had an update recently. They just had a a new update actually, while I was filming this. There you go, we're updated. Pretty easy. So, these are the top 10 slash commands and these are in-chat commands, so let's go back to the chat.
And that's just Hermes chat, right? There we go, back to the the terminal. Still with Mimo V2 Pro still. So this is my idea, here we go, I had an idea to start a business that uses AI to make custom children's books. Instead of just the name, it can use AI to create custom illustrations in the style the customer requests, as well as story themes, etc. Please do a thorough research into other companies and competitors who may provide this service, report back to me with their names, and try to identify any market gaps that may exist.
You can see it's initializing the agent here. And it's preparing the web search to find. Got some good leads. You can see these are all the tool calls it's, it has. I think this is a good business idea, it's a real one I had. Hopefully we can find some kind of market gap. Let me search for a few more angles to round out the competitive picture.
Uh, okay, and here we go, we got it our report back, pretty quick, as you just saw. Um here's a thorough breakdown of the competitive landscape and market gaps. Competitive landscape, market falls into three tiers. I have organized them from legacy players to direct AI competitors. The traditional, personalized book companies which were pre-AI.
So let me see the AI-powered ones. Here we go, direct competitors. These are the companies doing what you describe, using AI for both story and illustration generation. This is your actual competitive set. Uh, there's a lot. This might be a saturated market. Might not be workable.
You can see here, just that quick search it had found a lot of competitors. Market gaps and opportunities, there are some, based on competitive analysis, here are the underserved areas. Custom art style selection. So most competitors let you customize the character, but you lock you into one illustration style. Few let the customer choose or define the actual art style. The user said, "in the style the customer requests," this is the real gap. That could be something, right?
Older age groups, almost everyone targets ages 2 to 7. Uh, I really want children's like baby books. Those are the ones everyone's buying as like gifts. Bilingual, multilingual story generation, it found a lot of uh gaps here, but I don't know if any of these are workable for me. Audio first, culturally specific, author creator tools that's more on the B2B side.
So here we go. Wonderbly, I see me. So there's actually a lot. So, it gave it my my big take. The biggest white space is custom art style control combined with therapeutic niche story themes and older age groups. Most competitors are fighting over the 2-year old picture book market with similar AI illustration capabilities. If you let a parent say, "I want a watercolor style chapter book for my 10-year-old about coping with moving to a new school," that product does not exist today. The bilingual angle is also a massive untapped market with very little competition.
So there we go. There's your kind of market report, competitor report, just for this first video we did something simple, but I think this is a a task that a lot of people have, right? We all have random business ideas. Or maybe it's just me. And you can use your Hermes Agent to check on them, check on the competition, see what kind of market gaps exist, and try to plan when you want to start a business.
And there you have it, that is the end of Module 1 of my Hermes Agent masterclass. This one updated for July 2026. And now you should have Hermes installed, a provider configured, we learned about the top CLI commands and top slash commands, uh we did our first real artifact, we did research and developed this competitor report, and we have a memory file that just started to learn about us.
So we've done a lot, but we're just getting started with Hermes Agent. Module 2, uh which is going to be on deployment, different ways to deploy, on- uh deploying backends on Docker, uh as well as setting up messaging apps, so that you can actually talk to your Hermes agent on your phone through Telegram or Discord. So I hope you enjoyed this video. All 10 masterclass videos are now available on YouTube, but that's going to be it for this video, hope to see you in the next one.
Video
Video
