Hermes Agent Full Tutorial INSTALLATION + USECASES

Added: 2026-05-15

[00:00:00]A few months ago, I made a couple of videos about Open Claw. I was genuinely into it, but then things started going sideways. Security flaws kept popping up with almost every release. There were some pretty wild incidents where agents were doing things they really shouldn't, and the community started getting more and more divided on it. So, when I heard a new competitor just dropped, one that's more stable, more complete, and actually has a self-improvement loop built in, I had to check it out. That's Hermes Agent, and today we're installing it and setting it up the right way. Oh, and quickly before we start, I post three weekly. So, subscribe and hit that bell to become a code head and get notified as soon as I drop. All right, let's go. So, first, what even is this thing? Hermes Agent was built by Nous Research. It dropped in February 2026 and crossed 140,000 GitHub stars in under 3 months. But, here's the thing I want to be clear about up front. Hermes is not a coding copilot. It's not a chatbot you open when you have a question. Think of it more like an employee. Every conversation you have with it, it remembers. Every hard problem it works through, it writes down as a reusable skill, so it doesn't have to figure it out again. And you don't need to be running on your laptop to do it. You can be on your phone, texting it on Telegram, while it's executing tasks on a cloud server you haven't SSH'd into in a week. So, installation. This is the easiest part. On Linux, Mac, or WSL2, it's a single curl command. Paste it in, it installs everything automatically, no manual setup, no dependencies to chase down. But, if you've not been living under a rock for the past few months, then you know how dangerous it could be to host something like Open Claw or Hermes Agent directly on your main machine. But, buying an extra machine or spinning up a virtual one can be very expensive and resource-intensive. That's why today we will be hosting our Hermes Agent on Hostinger, the sponsor of today's video. Just click my link in the description, choose a plan, I recommend the KVM 2 for the best bang for your buck, pick your region, skip the manual Linux setup, and deploy Hermes agent in one click using Hostinger's pre-built Docker template. Set your environment variables, confirm your plan, and you're good to go. Here's what makes it the best choice. Hermes agent on Hostinger costs a fraction of hosted AI solutions.

[00:02:11]And unlike those, your API keys, learn skills, and conversation history stay completely on your own server. No throttling, no per agent fees, and your agent stays online 24/7, even when your laptop is off. Need more power as your agent grows? Scale your resources with a few clicks. No migration needed. Plans start at just a few dollars a month, and you can save even more with code codehead at checkout for an additional 10% off all yearly plans. Once it's running, you need to pick a model and select a messaging channel. Claude Opus is excellent. Gemini works well, too.

[00:02:46]But for this tutorial, I'm going with Qwen 3.6 plus, which has a context window of 1 million tokens. Here is how you can get it for free with Hermes agent. You first need to select the news portal option, which is usually number one. It will give you a link which will bring you to this page. Now, as you can see, you have the free option here. All you need to do is click the subscribe and connect button, and you're all set.

[00:03:10]As you can see, I have the Qwen 3.6 plus option for free in my Hermes terminal.

[00:03:15]Or alternatively, you can use Owl Alpha in OpenRouter, which, unlike Qwen 3.6 plus, which will not always be free, this model is truly completely free.

[00:03:25]Just grab your OpenRouter API key and select it into Hermes setup page. Then you have to go through some basic config. Just keep clicking next until it asks you to connect a messaging platform. Select setup messaging now and select the Telegram option, which is usually the first one. Okay, now after selecting Telegram, we need to create a Telegram bot. For that, we need to head to the BotFather and type the command newbot. It will ask us what we want to name our bot, and it will give us back a token once the bot is created. We just need to paste that token back in the Hermes agent setup page. After that, send the message to the bot, and it will give you a command to pair it to your Hermes agent. Just paste it in the terminal, and the connection will be approved, and we're good to go. Okay, so you've installed it. Most people at this point to start chatting and then wonder why it feels like any other AI. Here's what you actually need to do before anything else. Step one is what I call a brain dump. You're going to tell Hermes everything about yourself. Your name, your job, your current projects, the tools you use day-to-day, how you like to work, what your goals are, and don't be brief about it. For example, here's what I told it about myself. My name is Code Head. During the day, I work as a full-stack software engineer. My real home is back-end. I tolerate front-ends enough to get by. At night, I run a YouTube channel where I make short, punchy videos about useful dev tools, what's actually happening in software engineering and AI right now, and human side of being an engineer in 2025. The burnout, the identity stuff, the weird relationship we have with our own craft as AI starts eating into it. Step two is what I think is the most underrated thing you can do with any AI agent. Once you've done the brain dump, you flip it around. You ask Hermes what it thinks.

[00:05:08]The exact prompts I use is something like, "Based on what you know about me, what are the best ways we can work together? What use cases would actually help me? Ask me anything you want to know." And then you just let it go. What happens here is really useful. Instead of you sitting there trying to come up with things to use it for, the agent surfaces ideas based on your specific situation. When I did this, it came back with suggestions I hadn't thought of. It also asked me follow-up questions that made me realize I hadn't explained certain things clearly, which made the memory even richer. It's a small thing, but it changes how you think about the whole tool. Now, let me show you the use cases I personally set up. Take these and adapt them to your own life. Don't just copy and paste. The first one is a morning briefing. Every day at 7:00 a.m., Hermes sends me a summary of what's happening in tech and AI, but not just a generic news roundup. It knows I'm a software engineer and I make YouTube content about tech. So, it filters specifically for tools and updates that would matter to those two things. New libraries, model releases, workflow tools, things I might want to cover in a video. The second one is a daily check-in. Every morning, Hermes asks me what I'm working on. Dev tasks, video ideas, whatever I've got going on.

[00:06:19]I give it a quick rundown, it updates its memory, and then it'll sometimes come back with something like, "The thing you described doing manually every Thursday could be automated. Here's how." Or flag a tool that's directly relevant to something I mentioned. It's not always useful, but when it is, it's actually useful. These are mine. The reverse prompting step I walked you through earlier, that's where your version of these comes from. Don't just copy what I set up. Tell Hermes who you are, ask it what it would do for someone like you, and let it show you. One last thing before the outro. Tell Hermes to set up a skill check cron job.

[00:06:53]Basically, you're asking it to periodically go online, look for new community-built skills, and suggest the most relevant ones based on what it knows about you. Hermes skills are built on an open standard called agentskills.io, and the ecosystem is growing fast. New skills get added regularly. If you don't set this up, you'd have to manually check for updates yourself. With this running, your agent scouts for new capabilities on its own and filters for what actually fits your workflow. That's the self-improvement loop doing what it's supposed to do. Not just learning from your conversations, but actively keeping itself up-to-date. Most AI tools are exactly the same every time you open them. Hermes is different in one specific way. It accumulates. Give it a month and it'll know your workflow better than tools you've spent years configuring. The install link is in the description. The model recommendations and the prompts I used for each use case are in there, too. If you got something out of it, you know what to do. See you in the next one.