← Back to search
Hermes Agent: From Setup to 24/7 AI Assistant (Complete Guide)
Leon van Zyl · 2026-10-30 · 26 min
Show full episode description
Learn more about your ad choices. Visit megaphone.fm/adchoices
✨ Episode Outline — click any point to jump to it in the episode
Problem solved
How to install
Hermes Agent (terminal or new desktop app), connect a model provider, and run it 24/7 reachable from web,
Telegram and
WhatsApp.
Benefits
- Desktop app gives a ChatGPT-style UI with conversations, projects, file access and terminal
- Persistent memory keeps preferences and workflows across chats
- Agent creates reusable skills from workflows you teach it
- Scheduled automations like a daily morning research report
- Telegram bot locked to your user ID so only you can chat
Use cases
- Connected Hermes to Ollama via custom endpoint; local Qwen 3.6 replied after ~2 minutes first load
- Signed into OpenAI via OAuth (Codex) and selected GPT 5.6 with high effort, thinking on, fast mode off
- Set up a Telegram bot via BotFather, restricted by allowed user IDs
- Taught workflow: log into system, pull stats, add to spreadsheet, email it — saved as a skill
- Enabled extra tools (Home Assistant), MCP servers, and browsed community skills in the hub
KPIs / results
- First local Qwen 3.6 response took almost 2 minutes to load into GPU memory
- Provider option 32 = custom endpoint for LM Studio/Ollama
- Only one Ollama model installed on the demo machine
Hermes Agent received a ton of updates since my last video, from a new desktop app, a web interface, and a way more streamlined setup process. And when you combine it with OpenAI's GPT 5.6 sole model, it's an absolute beast. If you're new to Hermes Agent, it's this really powerful AI Assistant that grows and learns from you based on your conversations with it. So it can learn about you, about your preferences, your workflows, your business. It can even connect to third-party platforms, so it could read your emails, manage your calendar, and do a lot, lot more. It's even able to create skills to keep up with you. In this video, we'll go over everything from setting up Hermes Desktop on your own machine and connecting it to a provider like OpenAI. And you do not have to use OpenAI in this video, feel free to connect it to whatever provider you want following these exact same steps. I will then show you how you can get Hermes to run 24-7, even if you power off your PC. This will also allow you to access Hermes Agent from a web interface and from your mobile phone using something like Telegram and WhatsApp. By the end of this video, you'll have your very own AI Assistant that runs 24-7 and that you can access from anywhere. So go ahead and bookmark this video, as this is going to be the Ultimate Hermes Agent Guide. Alright, to get started, go over to this website, which I'll link to in the description of this video. This is the official Hermes Agent website and if you want, you can read through this and get an understanding of what Hermes Agent is all about. But in a nutshell, this agent lives everywhere, so you can access it from your PC, from the web, you can access it through all sorts of channels like Discord, Telegram, WhatsApp, email, really whatever you want. You can hook it into your agent. It also uses persistent memory. So that is how the agent learns about you. So as you're sharing very important details about your preferences or workflows, it's going to store that information in persistent memory. So the next time you chat to the agent, it will keep all of those learnings in mind. And Hermes Agent also focuses on automation. This means you can ask it to set up some kind of recurring task or run something on a schedule. So maybe every morning you want some kind of good morning report where the agent needs to go online, perform some deep research or produce some report, maybe by pulling information from some ticketing system. You can do whatever you want and the agent will be able to set up those recurring jobs and automations for you. Hermes Agent is stupidly popular for a reason and that's really why. Now we've got a lot to go through in this video, but first let's focus on setting up Hermes Agent on our local machine. Here we basically have two choices. We can set up Hermes Agent in the terminal or we can install the desktop app. So for completeness sake, let me show you the terminal setup first and then we'll have a look at setting up the desktop app. In order to set up Hermes Agent in the terminal, simply go to this install via terminal section. And if you're using Mac or Linux, simply copy this command. Or if you're using Windows, just copy this command. Since I'm using Windows, I'll copy this. Then I'll open up PowerShell. Then it's based in that command and run this. This is going to download Hermes Agent and install all of the dependencies. Then you will be asked to connect a provider. Now, as you can see from this list, there are tons of providers from NALS portal, which is kind of their official portal, where you can also access a whole bunch of models and you pay their subscription fee. We've got access to Open Router, OpenAI, Anthropic, Quinn. There are just so many different options here. A super fun one is the custom endpoint, where if you do want to run local models using LM Studio or Olama, you can actually just use custom endpoint and then use free models on your own machine. In fact, as a little bonus, let's actually do that. So I'll enter 32, which corresponds with custom. So let's just enter that. Then I'm going to paste in the endpoint to Olama. For the API key, that's usually just the word Olama. Then for this, I'll just say yes. And this is just saying that Olama is actually set up on my machine and it found one model. And by the way, if I just have a look at my list of Olama models, I do indeed only have one model downloaded. Now, if you're new to Olama and you also want to run models locally on your machine, I'll link to a video and I'll show a card on the screen right now where I've got a dedicated video on setting up Olama and local models. But don't worry if you can't run models on your own machine. For the remainder of this tutorial, we will be using OpenAI and I'll show you how to connect it to other models as well. So for this, I'll just say auto detect. It's saying that it detected QIN 3.6 latest. So do I want to use that model? I'll just say yes. And I'll just press enter. And then where do we want to run the terminal? Backend, local, Docker or whatever. I'll just go with whatever the default is. But then we can configure messaging platforms if we want. So we can hook up things like Discord, email, WhatsApp, Telegram, you name it. I'm not going to set up any of these at the moment. We will do that later on in the tutorial. Then for the Hermes tool configuration, I'll just go with whatever they have available by default. Then let's enter three and that should be it. Let's run the command Hermes and that will bring up this Hermes agent CLI. And we can now use Hermes agent in our CLI tool. Let's just give it a try like A. And the very first time I run this, it is going to be slightly slow because it's loading this screen 3.6 model into my GPU's memory. And cool. I actually just received a message back from the model, which means the local model is working. It took quite a while. It almost took like two minutes to get that first respond back. So I sent it the message, are you stuck? And this time it actually responded way faster. Either way, that is how you can spin up the terminal and use local models. One of the new features from Hermes agent is actually the desktop app. And this is really cool. From the same website, it should show a download button for your operating system. Since I'm on Windows, I'll simply download this and run the installer. And I'll just click on install Hermes. And again, this will download and install all of these different dependencies. And after installation completes, we can go ahead and launch Hermes desktop. And it will look something like this. This looks very similar to every other chat interface, right? Like chat GPT or Claude. We can see our conversations on the left and we can organize them by, you know, groups or by projects. We can chat to the agent as well. And in the chat window, we do have access to all of our chat models. Now, I do want to mention that I have access to the QIN 3.6 model here because I went through that local CLI process. At the time of recording, there's no way to set custom endpoints directly in the desktop app. So if you do want to use local models running on LM Studio or Olama, you do have to go through that terminal process up front. So if that's something you want to do and you don't find any way within this add to set custom endpoints, that is really the solution. So at the moment, I can, of course, select my local model and let's just say, hey, just to prove that it actually works. And there we go. We get the response back. And this is coming from my local model. Now, also, when I click on this dropdown, I can see a whole bunch of different models as well, like these Gemini models. And I do want to mention that the only reason I can see Gemini right here is because I am signed into my Gemini CLI tool. So for you, this list might be completely empty and you can't interact with any models yet. But please don't stress. Let me show you how to connect to a provider. All right. In order to connect to a provider, go to settings up here in the top right corner, then go to providers. And from this list, you might not see any providers connected at the moment, which is perfectly fine. Hermes agent will try to look for any other CLI tools or other coding agents that you might be signed into already. Like it detected that I've got Claude Code. So it's just kind of using this Claude Code subscription, which I would not recommend. If you're not aware, there's a lot of bad blood between Anthropic and these AI assistants like OpenAI and Hermes. And Anthropic really wants you to pay for your own usage and not piggyback on your subscriptions. And I do not recommend using API keys with these agents. It's going to be way too expensive. Look, if that's something you do want to do, there's this API keys option here as well. And you can see all of the different providers here. So, for example, the new Kimi 3 model just came out. So maybe you want to connect this to Kimi. You can just go sort of to the Kimi option here and connect it to Kimi or Moonshot or whatever you want. But for this video, I am going to use OpenAI and GPT 5.6. So what we can do is connect to another provider. Let's look for OpenAI OAuth. Then I'm going to select my OpenAI account. Let's pick on continue. Then this is going to ask us for a code. So let's just copy this code and paste it into this field. Cool. Let's continue. And now we're signed into Codex. And now we can see that OpenAI is connected. This means if we go back to our chat, we can click on this dropdown. I can already see the GPT 5.6 models. But if you don't see anything, just click on refresh models. And then I'm going to select GPT 5.6 soul. And there's another thing. When you mouse over any of these entries, you can see additional options. Like we can enable or disable thinking mode. This simply means agent will spend a little bit more time thinking about its response before giving you an answer. You can also enable fast mode, which is going to be a bit more expensive, but you're going to get faster results. I'll just switch that off though. Then we can see our effort level as well. I prefer high. And that's really it. So now we can start a new conversation. And let's say, hey, how are you? And let's send this. And that was blazingly fast, but we got a response back. You also notice on the right hand side that this agent does have access to your file system. So you can get it to actually do everything that the coding agent can do. It can dive into a project folder, work on code changes, create new projects for you, whatever you want. You also have access to an integrated terminal. So if you ever wanted to run terminal commands, you can do that as well. Let's do something like echo. Hello. Let's send this. All right. And cool. We get the message back. Hello. In the left hand side, we can see all of our conversations. We can also do things like pin conversations that will kind of stick them to the top. So we can access them really easily. We can rename conversations. I'll just call this hello world. So all the stuff you would typically expect to be able to do with these chat interfaces. One of the really core benefits of using Hermes agent is it's really good at managing skills, working with skills and even creating its own skills. So if you teach your Hermes agent a certain workflow that you go through every morning, where you log into a system, you pull certain stats, you then add those figures or numbers to some kind of spreadsheet, and then you email that off to somewhere else. You can teach that workflow to your Hermes agent. And that agent will create a new skill that we'll use going forward. And that skill will also show up in this list. You can also assign additional tools. So some of these tools are currently disabled, like this home assistant. If you want, we can enable this tool. And if we click on it, we can go ahead and configure whatever that tool needs. We can also go to MCP. So if you're familiar with using MCP servers, you can easily add your own MCP servers using this menu as well. And what's really cool is if you want to search for additional skills or tools or MCP servers, you can click on browse hub. And from here, you can search for anything you want. So I don't know, let's see if there's any like YouTube relevant skills. And there are plenty of skills. And you'll notice that some of them were built by the community. Then if we go to messaging, we can connect our agent to external chat applications, like Telegram or WhatsApp. We've got Slack, Discord. There are so many different ways to integrate your agent into these chat apps. And yes, you can have more than one enabled at the same time. So maybe when I'm out and about, I want to use WhatsApp. Or if I'm at my PC or on some work computer or some kind of VPS, I might want to use Slack. You have plenty of options here. For each of these, you do have to just configure it. So if you want to use the Telegram bot, we will have to just provide our Telegram API key. In fact, let's actually do that. So in Telegram, I'll simply go to Botfather and you can do the same. Look, installing Telegram is really easy. Just go to telegram.com, download the app, sign in, then just search for Botfather. It's just the contact in the list. And then you can type the word slash new bot. And that will take you through a wizard of setting up a new bot. By the end of it, you'll get this API key. So I'm just going to copy this API key and add it to this field over here. Then for allow Telegram user IDs, what you can do is look for this contact, user info, and you can then just run the start command and that will give you your user ID. This is needed so that other users can't chat to your bot as well. So this will just kind of lock it down. So you are the only person who is allowed to chat to your bot. If you want to add additional users, so maybe you want your partner or your children or your colleagues to access the bot as well. You can simply add them by entering a comma and then adding their user IDs to this list as well. So I think that's it. Let's actually save these changes. Then I'm just going to open my bot. Let's say, hey, you will see this message saying that the messaging gateway stopped. So what you have to do is just go down to the bottom down here. Then let's click on this restart gateway button. Right now, Telegram is connected. So what we can do is just go back to Telegram. Let's say, hey, I'm just going to click on this set home button. All right. Let's say, how are you doing? All right. So we got a response back and check this out. In Hermes desktop, we can see this section down here called Telegram. Telegram and from this channel, we can view all of the new messages. And at this stage, we only have one message. So this just confirms that Telegram is now chatting to our agent running on our own machine. Now, this is really cool, but there's one major limitation. At the moment, the agent is running on our own machine. So if we ever switch our PC off, the agent is going to go offline as well. And we can't access it from Telegram anymore. In order to run this agent 24-7 without it ever going down, we need to deploy it to a VPS or a virtual private computer. What we'll basically do is rent a computer for super cheap from an external service provider. And we'll run Hermes agent on that machine. What we'll then do is connect our Hermes desktop and Telegram to that instance of Hermes agent. That means we can switch off our PC and we'll still be able to access Hermes agent from our phone and from a really cool web interface. I can't wait to show you. Let's dive in. Now, there are many different ways to deploy Hermes agent to a virtual private server or a VPS for short. Now, personally, I don't like messing around with infrastructure. I personally hate it. And I like to keep things as simple as possible. So from that point of view, I think Hostinger is the easiest solution. It's simply a one-click install. I also partnered with him for this video to give you an additional 10% off. Now, if you do want to follow along, you can simply go to this URL, which I'll link to in the description of this video. Then you can choose a plan like KVM1, which would be perfectly fine for Hermes agent as well. I'll go with KVM2. So let's choose a plan. Then you can choose your period from one month, 12 months, 24 months. I'll just go with 12 months. Then for this section, I'm actually going to disable ready to use AI. It's really not necessary. I'm going to disable all of these options. You can also choose your server location. So try to find a location that's close to where you live. And then don't change anything under operating system. Then under operating system, let's search for Hermes agent and ensure that this one is selected. Then under order summary, go to have a coupon code and enter Leon Hermes and click on apply. Now I've already used that code, but this will give you an additional 10% off. Now during the setup process, you will be provided a username and password. If you ever forget those credentials, please don't stress. All you have to do is go to Docker manager. Then under projects, search for this first record over year and go to manage. Then scroll down to environment. And you can simply grab your username and password from here or change it from here as well. Either way, let's go back to Docker manager and projects. And then under this Hermes agent entry over year, let's click on open. Then let's enter that username and password like so. Let's click on sign in. And how cool is that? You can actually go ahead and bookmark this URL because you can access this URL from your phone or anything. So it's a public URL that you can access from anywhere. It looks and feels very similar to the desktop app. We can create chats, we can view sessions, but it gives us a lot of other capabilities as well. We can view all of the Chrome jobs and those are all of those little scheduled jobs that the agent can set up and all of the little automations. We can also view all of the skills available to this agent as well. So from the skills tab, we can add skills, we can disable or enable skills. We can go to plugins and of course, view all of the install plugins, disable them, everything that you would expect to do with the agent on your local machine. Same thing with the MCP servers. We can also connect the channels to it like telegram and a lot, lot more. So let's start with the basics. When we click on the chat window, we have to just set up a provider. So in order to do the setup, let's click on front slash and enter the word setup. Then we have a few options. We can use quick setup, which will connect to NALS portal. But I actually don't want to do that. I want to use chat GPT and my chat GPT subscription. So I'll just go with full setup. From the list of providers, I'll just go to OpenAI and I'll select OpenAI Codex. Then I need to open this URL and we also have to copy this code. So I'll just go to my account, continue. Let's paste in that code and continue. Cool. Now let's select our default model. I'll go with GPT 5.6 sole and I'll just go with whatever the default options are. Right. So then we can configure our channel. And for this, I will use Telegram. Note you can select more than one provider as well, but I'll just keep it simple and set up Telegram for now. You can easily set up additional providers in the future, simply by asking the agent to help you set up whatever channel you want to configure. Then for the setup process, I'll just go with number two, which is manual. Then we have to provide our Telegram bot token. I'm just going to reuse the same bot as from before. So I'll paste in that token. Actually, I seem to be running into a bit of an issue here. So I'm going to press Ctrl C just to go back. I'm going to show you a different way to set all of this up. So if you run into issues as well, this is another cool way to do it. So I'll just authenticate using OpenAI again. I'll go with those existing credentials. Let's select GPT 5.6. I'll just go with the default for the messaging platforms. I'm just going to click on enter without selecting anything. Then for the different tools, I'll just go to done. And let's just see if everything works. I'll just say, Hey, and we get a response back from the agent. So now I'm just going to ask the agent to configure the channel for me. Please help me configure Telegram, please. The token is. Let me just paste in the token. The user ID is then back in Telegram. Let's go to this ID. Contact. Let's copy my ID. Then I'll paste in the ID and press enter. All right, cool. So I just asked the agent to send a test message to Telegram. And we did receive a response as well. Just say howdy back. And Hermes agent is typing. And of course, we get our response. Right, cool. This means we now have an agent that can run 24 seven on a VPS. And we can call it from Telegram or whatever channel we set up. However, the desktop app is still pointing at its own sort of thing. We've connected it to OpenAI, but it's not connecting to our main agent that's running on the VPS. So how can we ensure that the desktop app can connect to our remote Hermes agent? Well, thankfully, that is really simple. The first thing we need to do is disconnect Telegram from this local instance. Because at the moment, Telegram is just kind of being used by both the desktop app and the agent running on the VPS. And we want the VPS version of it to be the single source of truth. That is our main agent. So all we have to do is go back to messaging. Let's go to Telegram. And let's just disable Telegram over here. Cool. And I'm also going to restart this local gateway. And that will just ensure that Hermes desktop is no longer listening for Telegram messages. Then the second thing we need to do is go to settings. Then let's go to gateway. At the moment, we're using the local gateway. But what we want to do instead is point to our remote gateway. Then what we can do is grab the URL. So basically this entire address over here, excluding chat. Then let's paste it into this field over here. Let's click on sign in. This will ask us for our remote Hermes username and password. I believe the username was Hermes. And I'll paste in my password. And let's sign in. And now this is saying that we are currently signed in. Let's click on test remote. And now we are connected to our remote Hermes agent. So let's just close this. And I just actually restarted Hermes desktop as well, just to make sure all the changes do take effect. Now, this instance of Hermes agent should be synced with the agent running in the cloud or on our VPS. Let's just ask it. What is the current weather like in New York? So just a very specific question. Let's send us. All right. Our agent is currently thinking. Let's run this command. Right. So we get the response back that it's 75 degrees Fahrenheit in New York. If we go back to our web session and we go to history, we can see that same message over here. Where we asked the agent what is the current weather in New York? So this is all synchronized now. In fact, let's try to send a new message from this chat over here. In fact, let's go to chat. Let's create a new chat. Let's say please set up a cron job that runs every morning at 9am to give me the latest news from Claude and OpenAI. Cool. Let's send this. This is asking us where these messages should be sent to. I'll just say telegram. So basically every morning at 9am, I should receive a report to telegram giving me the latest news from OpenAI and Anthropic. Right. Cool. So it's set up this little cron job. If we go to cron, we can view our daily news cron job over here. And of course, if we want, we can go into edit mode and we can change the schedule ourselves or just ask the agent to do it for us. What's really cool as well is if we go back to the Hermes desktop app, have a look at this. In the bottom left corner, we now have this cron job section over here and we can click on manage and see exactly what the settings for this job is. So we can see the frequency when it's going to run next, the actual prompt that it's going to use. It's all there. And we can manually trigger it from here as well. And we can set up new crons. This just proves that everything is indeed connected. Or let's actually go back to telegram and it say, please could you list all of my scheduled tasks? Let's send this. All right, cool. So it's replying and just having a look at this, we can view our scheduled tasks in here. Please could you change the time to 7am instead? All right, cool. The agent is saying it rescheduled the job. And if we actually refresh this cron job list, this has been changed to 7am. How cool is that? Now, finally, there are just so many things we can do here. We can hook this up to Gmail, to Calendar, to whatever system we want. Hermes Agent is a lot of fun to use. If you haven't tried it, do you want to give it a try? It is so much fun. If you enjoyed this video, hit the like button and subscribe to my channel for more agentic coding tutorials. If you would like to learn how to use coding agents efficiently, I actually have a completely free course. It's called the 7-Day Builder Challenge. This teaches you the RAMP framework. This is the framework I use with all coding agents and even free coding agents, by the way. And by the end of it, you'll build your very first project. I'll leave a link to the community in the description of this video. Thanks again for watching. I'll see you in the next one. Bye-bye.