HERMES AGENT SETUP: the OpenClaw killer is here

Recently in news research released Hermes agent and in some very important ways it's better than open clock. The thing that immediately stood out to me about the Hermes agent was this idea that it keeps getting more capable. The longer you run it, the better it gets. It has a persistent memory [clears throat] and auto-generated skills. Each skill that it develops it treats almost like a the scientific project. Form hypotheses, test them out, see if you can improve your abilities at the thing that you're doing. It does this for your projects. It learns your projects and it never forgets how to solve the problem. By the way, I'm so happy that News Research is the company behind this and yes, that's how you say the name. It's News. A lot of people are getting a very fancy saying it, but it's just News Research. We had a chance to interview Mephisto aka Karan himself. He does the model behavior at News Research and the interview we've had with him was pretty mind-blowing if I may say so myself. So, the fact that it's their AI agent that's really blowing up right now just makes me very very happy. It is a great team. It's a great lab and I'm very happy to see them succeed. This is a distributed training lab. This is an open source lab and their whole idea, their goal comes from this idea that these big AI labs should have no business setting the morality and ethics of your AI model. Those should be neutral. The user should have the power to decide how the model behaves, not the big AI labs. By the way, if you recall World Sim, maybe you've heard of it at some point that that's them. So, I've been meaning to do a video about Hermes agent because it's blowing up. A lot of people are talking about it. This idea of self-improvement, of continuous iterations. I think it definitely captured a lot of people's imaginations. So, today was the first day I was able to set up and I've been experimenting with it. So, I know I'm a a little bit late to the party, but if you haven't installed yet, if you haven't tried it, I hope today will be the day that you do. You don't have to be a developer, you don't have to be super tech savvy. A lot of stuff is fairly easy and getting much easier with the assistance of your favorite AI chatbots. It's also important to note that the Hermes agent is going to power their agentic reinforcement learning pipeline, expanding Atropos so you can run reinforcement learning with Hermes agent primitives and it supports mass scale data generation out of the box. We're not even going to talk about that today because the scale of this thing is insane. The idea is to have a truly scalable asynchronous large language model reinforcement learning. Their vision, once you begin it to grasp it, is kind of staggering. They're already building in Karpathy's LM Wiki. It's now a built-in skill within Hermes agents. Here's the thing, Hermes agents and even the entire company, Nous Research, they're really about local open source models. That's what they live and breathe. So, expect to see more like this where Hermes agent is running locally with Gemma 4 on a MacBook Air with Turbo Quant Cache. If you recall, we covered that, I don't know, less than a week ago, I think. And then just yesterday they released their version 0.9.0, the everywhere release. This thing is blowing up. Don't miss it. If you missed Openclaw, jump on this train. Nous Research isn't paying me to say this. I don't have any financial relationship with them. I love the fact that Openclaw was built by a solo dev. And I love that the next big thing it seems to be coming from this particular lab. So, here is Hermes agent version 0.9 release. It's kind of exciting. I've only had a few hours of really testing it, but I am so far very excited. Let me show you what it comes packaged with from day one. in mind as it learns more about you, your projects, what you're trying to do, it's going to start building these jobs and these skills and every day trying to improve them to make them better. It's that do, learn, improve loop that it's going to keep going through on your behalf. So, those self-improving skills, we're not even looking at that yet. This is just what it comes with standard. Right now, fresh install, we have 74 installed skills. Research academic papers from archive, you you create article, create a video, audio craft, and it can interact with Claude code built in. You do need the Claude CLI installed on the same computer, but as it's doing its thing, it's able to delegate tasks to Claude code. Remember how Anthropic said you can't use the OAuth to power, let's say, Open Claw, but you can use it for Claude code. This has a skill to delegate tasks to Claude code. So, I don't know, could be huge. And by the way, Codex as well. Tons of skills for local models, quantization, fine-tuning LLMs, a lot of various get help stuff. Look at this, it has a god mode skill. What do you think that is? Nazi 2 jailbreak API served LLMs using various god mode and parcel tongue techniques. You have a Hermes native pre-fill system of prompt integration. So, if you wanted to bypass some safety filters, you have this skill sitting there. You have a various reinforcement learning training skills, Llama local models, you have Notion, Obsidian, Polymarket, PyTorch, Segment Anything, Stable Diffusion, you get Whispers, the OpenAI's text-to-speech model, as well as a deep research on a Twitter and YouTube. There's a lot here, and we're just scratching the surface. I think this team is going to keep building and keep shipping. And I wouldn't be surprised that over the next however many months, all the big AI labs will try to copy whatever they can from this. There are hundreds of models to choose from. Open Router is one of the default providers. So, you can have your Claude Opus 4.6 or switch to Grok 420 or Nvidia's Nemotron or the RC's AI Trinity models, right, which are notice they're free. Some of the Nemotron and Trinity models are free right now. According to Pinch Bench, which is a benchmark for testing how good different models are at completing actual Open Claw tasks, notice the RCAI, the Trinity model, is oh so close to Claude Opus 4.6. So, I really want to try it out because if it's as good as this shows, that means you can get the power of Claude Opus of OpenAI's GPT 5.4, it's better than GPT 5.4, but in a model that is much, much cheaper. This one is the Trinity large thinking. That's not the free one, but look at the pricing. Not sure how well you can see that, but it's 22 cents for the Trinity model versus $5 for Cloud Opus 4.6, 85 cents for the outputs, and $25 for Opus 4.6. Anyways, there's a lot to be excited about here. This, I believe, will be a longer term obsession for me. So, definitely expect more videos. But today, I just wanted to help people get up and running in the easiest way possible. You don't have to be a developer, you don't have to be an engineer. You need to be able to interact with, you know, your favorite AI chatbot, whether it's Claude or OpenAI or Gemini or Grok, and be able to put stuff into it, ask it questions, and then follow instructions. If you can do that, you can have this thing installed and running within the hour. The whole thing is open source. It's free. You do have to pay Most of these models, you do have to pay for the API costs, but if you love Open Claude, you definitely should try this. If you missed out on Open Claude, this is your way to really jump to where the action is right now. So, today, let's look at the installation. But expect a lot more guides about how to use this in combination with Open Claude, and also how to really take advantage of their self-improvement loops. And this idea of integrated reinforcement learning training within the Hermes agent. I don't know, I'm getting excited. So, please make sure you're subscribed, hit the thumbs up. I do want to get this to as many people as possible. So, strap in, lock in, and let's go. All right, so to begin, we need a place to install Hermes agent. We can do this on a local machine like a desktop, some old laptop, a Mac mini, or maybe even one of those mini PCs. But judging from the previous videos that I did, what a lot of you are interested in is number one, making it very simple to install, and number two, making sure that it's very secure. And that brings us to our second option, which is installing it on a VPS, a virtual private server. As you can see here, News Research, the makers of Hermes agent themselves, recommend that as one of the best options. Installing this agent on a VPS comes with a lot of advantages. That means your agent is on a machine that's always available. It's always online and you don't have to deal with the headaches of maintaining local hardware. So, that's what we're going to use to install Hermes on it today. I'll be using Hostinger, who is the sponsor of this segment of the video. You already know that name. We've used them in many previous videos before. I know some of you get a little bit grumpy when sponsors are mentioned. Here's the important thing to understand. Hostinger is what I personally use. Here's my account. I use the KVM 2 service. When I host AI agents online, I use at this point exclusively Hostinger. And this is not an affiliate deal. I don't get paid when you sign up. Whenever I do a tutorial video like this, I just reach out to Hostinger and say, "Hey, you want to sponsor a video?" I'm not incentivized to get you to buy anything. I just show you the product that I use personally. And Hostinger gets to be mentioned in front of the very audience they're trying to reach. I'm talking about you, the smart and handsome AI enthusiast. Hostinger wants to be the main provider for us, the people here in the AI space and the automation space. And they are, in a word, killing it. They're doing a great job. I did not know about this until I mentioned it to them that I'm planning to do a Hermes agent tutorial. It's sometimes a little bit difficult to walk people through the initial installation of these things. Depending on the person's tech background, it can get a pretty hairy. So, that's why I was very impressed with the fact that they have a special page and a one-click installation for Hermes agent specifically. The plan is the same one that I've mentioned to previously. Two vCPU cores is a good number of cores for running agents. It prevents them from freezing and running into resource issues. You get 8 gigs of RAM, which is more than enough for one agent with room to grow. And And important with this particular installation. I'll tell you why in just a second. And finally, you have 100 gigs of NVMe disk space. The NVMe speed ensures that those are productive scans where the agents read various files on a disk that they happen basically instantly. At $8.99 a month, it's a no-brainer, but we can even get you a discount on that. One thing that I really liked about what they did is they created a persistent Docker container. You don't need to know too much about this. You don't need to know how to set it up. It's already been done for you. And this persistent Docker volume ensures that the agents learned skills, memories, session history, and configuration survive container restarts and upgrades. Your agents become more capable over time without losing progress, which is one of the main things going for Hermes agent. That's one of the core features. Now, let me walk you through the installation, but if you're ready to install a Hermes, go to hostinger.com/hermeswesroth and use code wesroth for an additional discount on yearly plans. All right, but let's fire this thing up. So, we're using the KVM 2 plan. This is the plan that I use and always recommend. If you use my coupon code wesroth, you get an additional 10% off the price. Select your length of time and the price is locked in for that period of time. The Hermes agent auto deploys with your VPS. This is what you need and that comes standard with this plan. Select any additional options that you want. Make sure that the server location has the best latency for you. I'll just leave it as is. And uh let's go. Next, it'll ask you to create an account and then as soon as you're done, within 10 seconds or so, it gives you this screen. Hermes agent configuration. You need an open router API key, an Anthropic API key, an OpenAI API key, and a Telegram bot token. This is a super simple. They give you the actual URLs you need to go to right here. So, openrouter.ai/keys. It'll look something like this if If don't have an account, log in with Google or create an account. Open router is the one ring to rule them all, so to speak. Basically, it allows you with just one API key to control pretty much any model that you want. They have all of the models you're familiar with. They have the various featured agent. Notice Hermes agent this year. And once you create an account, or if you already have an account, this is the page that it'll take you to. Here you can create a various API keys. If you don't have one set up, click create, name it, add a credit limit, reset the limit every whatever you want, and do you want it to expire? So, we'll call this Hermes. We'll give it 100 bucks every, let's say, day to start. We can always adjust it later. Expiration, let's just to be safe, we'll set it to 90 days. So, it's going to expire on this day. Click create, and then you will see your API key. That's the last time you're going to see it because after that you're not going to be able to see it. So, copy, paste it to where you need it to go. For us, it's in the Open Router API key. And don't lose it. Make sure nobody sees it. And then you can do the same thing for Anthropic. You can do the same thing for Open AI. And then you need your Telegram bot token. To do this, you have to go see the Botfather. But, do not be alarmed. He's pretty friendly. So, open up Telegram on your phone or on your computer or wherever you have Telegram and find this user, Botfather, @Botfather. And he's got an offer you can't refuse. I'm not doing the accent. And he can help you create and manage Telegram bots. Basically, you just need to type in {slash} new bot. So, at the bottom here, we're going to say new bot. This creates a new bot. What are we going to call it? Call it whatever you want. This is the name that it's going to appear under its name. It can be anything. Doesn't have to be unique. I'm going to call mine Hermes agent. Extremely original. I know. The Godfather says, "Good." I'm not I'm not doing the accent. But, basically, next you have to select a username. This one has to end in bot. So, you can call it for example Hermes underscore bot, Hermes dash bot, or just Hermes bot. And once you put that in, you'll get basically a response that gives you your bot's API key. Keep that safe, copy and paste it into the screen where you put your API keys. So, in our Hermes agent configuration, that's the Telegram bot token. Boom. And then we hit deploy. You might get a little survey, but at this point you're basically in. Your VPS is set up and your agent is running. If you recall on that original page where hosting there has the Hermes agent setup, they do have a documentation that walks you through how to get started. So, if you scroll down, we deployed the Hermes agent and now we need to access the Hermes agent. Now, the platform automatically created and started the Hermes agent container along with the required services such as a traffic traffic. So, the first step is to use the browser terminal to connect to the server. So, from this page, this is the first page that you sort of land on once you complete those setup steps like I just showed you, you'll end up here. Here's your Hermes, here's the traffic, and we'll hit terminal. Terminal can seem intimidating if you haven't worked with it before, but trust me, you're going to learn to love it. Here's the saying that they've installed Ubuntu Linux, which has my stamp of approval. And you basically type in the commands here. The first command, they just write it out for you here. This is the command. A CD is change directory, and then the rest is the directory it's in. So, slash docker and then slash Hermes agent and this XXX is probably the name of our agent. If you ever have trouble navigating a terminal like this, just open up your favorite chatbot, tell them what you're trying to do, and they'll be easily able to walk you through it. They love this sort of stuff. And I'm sure this is where we might lose some people. Certainly, oftentimes when a tutorials are written online, there's a a of missing steps and people get very confused. So, really fast, let's just take a second to get familiar with this interface. This is a command line interface, a CLI, and you just type in whatever commands that you want to happen and they happen. So, really fast you understand that where you are. This is root at CRV and then some number. Root, that's the user, that's you. This is the administrator, the superuser, that's the person in charge of the account. This user has the permission to do anything. And they're at this server or this machine called SRV something something something. There's a few basic commands that you have to know to navigate your way around. The first one is a CD, change directory. So, we type in a CD space {slash} docker. Docker you can think of as a container into which we can kind of install whatever software or operating systems that we want. It's a way for us to kind of install a pre-determined environment so that it's the same across any computer that you take it to. So, if you imagine like a shipping container, shipping box, whether it's at your house or in a truck or on a crate or out on a freighter in a sea, what's inside that shipping container doesn't change. And so, this allows us to basically copy and paste them and basically give everybody copies so that when they open it, it's identical to every other one. So, that's docker. And then, we're going to continue, we're going to do {slash} hermes-agent. By the way, these are folders on this machine. So, oftentimes, if you start typing and hit tab, it'll auto-complete the name of the folder for you. So, here if I hit tab, it knows that it's {dash} 7uxi, which is just a specific code for that specific agent. We're going to hit enter. Now, notice we're still root at this SRV, but now our directory changed to {slash} docker {slash} hermes-agent. And a big point of this installation is that everything is within that docker container, so as soon as we enter that container, everything's going to be set up for us. How do we enter that container? We type in docker space compose. Docker compose basically means that we're going to have this docker with a sort of composition it. So, a certain system, certain software. It's all sort of pre-built for us. Has everything that we need. So, docker compose space and exact basically means that we're going to be running a command within an existing docker that was already set up for us. So, we're not starting a brand new one. It's already there. Next, we hit space and with all these commands, you can do what are called the flags. So, for example, -h often times is for help. If you want to see the readme file, the help file, -h usually is it. But here, we want two flags. The first one is -i and this basically means that it's interactive. That we're able to interact with it. We're able to type and click on stuff, whatever. And we also want -t which means it's a terminal. It runs like a terminal, so we're able to give it commands, etc. So, we actually just combine those two flags together by by typing in -it. So, that basically says that we want this thing to open as an interactive terminal. Next, we put a space and we type in hermes-agent. That's just the name of the docker container that we're entering. That's what the people that put it together for us at Hostinger, that's what they called it. It's called hermes-agent. And then finally, it's space /bin/bash. And that just means open a bash shell or give me a command line where I can just type stuff. And if you're wondering, how do you remember all this stuff? Truth is, mostly you don't and not unless you're doing this over and over again. Hostinger has documentations on how to get started. So, they actually walk you through it and they have all the commands right here. So, cd docker hermes, etc. And then this is the command that we just typed out. All right, so we typed it out and we hit enter. And notice kind of where we are changed a little bit. We were at this crv thing. That's basically the the host machine, the the main machine, the computer that we sort of logged into. And now we're at this numbers and letters. This basically means we're in the docker container now. So, it's almost like a separate machine within the machine in which we're going to operate. And on that Docker container within that machine, we're in the folder {slash} opt {slash} hermes. All right, we're almost there. Stick with me. Next, we just activate what's called a virtual environment. I just think of it as a very specific toolbox that we need over there. We don't need all the other tools, we just need these particular set of tools. This basically makes sure that there's no version conflicts. And whatever you install within that sort of environment [clears throat] it stays in that environment, doesn't conflict with the stuff outside of it. But, let me type out the command so you can see what goes into it. So, first of all is a source. Source means run this command and apply it to my session. Then, a space and then the path to the activation script, which is {slash} opt {slash} hermes {slash} {dot} venv {slash} bin {slash} activate. And that's it. That basically opens up that virtual environment. And this way, if you install anything within that environment, it's not going to conflict with anything else on the computer. So, we're going to hit enter. And notice now you have this hermes in parentheses in front of where you are. So, this basically means that this virtual environment is activated. If any of this doesn't make sense or doesn't work, use your favorite AI chatbot to have it explain it to you. Or if there's an error message, it's usually able to figure out what the cause is fairly quickly. But, if this seems really archaic and weird, trust me, it's not as complicated as it looks. And this is really it. You are now where you need to be. You're within the Docker container where everything's already installed. And you've opened up that virtual environment, so now everything's sort of you got all the tools set up that you need. You got your special little toolbox and you're ready to go. Literally, now you can just type in hermes and hit enter. At which point, there it is. Hermes opens up. This is a version 0.9.0 released within the last 24 hours. If you scroll down, you can start chatting with it here. The model that you're talking to is shown there. If you want to see a list of all the commands that are available to you within Hermes, type /help and then hit enter and it'll show you everything that's available, all the various commands that you might want to use within Hermes, but also towards the end you can see the skill commands. There are a lot of the skill commands, so this one came with 74 installed. Things like researching scientific papers, doing research on YouTube, on X/Twitter, audio generation. Some of it is actually directing with Claude code. You do have to have the Claude code as CLI installed, but you're able to just interact with it from here. Same thing with Codex. You've got tons of a machine learning stuff, GitHub code review, etc., etc. You have a god mode. This is for people that are fans of a Pliny the Prompter. This allows you to do some jailbreaking and the AI safety research. These are some of the open source jailbreaks that can be used to break down these models. You can use various things in the Google Workspace. This uses OAuth 2, which has to be set up. That's a little bit involved, but it's there. Hugging Face Jupyter Notebooks, and of course Minecraft because what would this be without Minecraft, really? Various other organizational apps. We have a Nano PDF, Notion. We have Obsidian over here, PowerPoint, PyTorch, everything you can think of. And of course, the ability to create your own or download others from the internet. These are the skills that make Hermes come to life. All right, so now in this kind of interface, when you're typing something here, Hermes will respond. I'm going to say hi. Notice it's initializing the agent and now Hermes is responding. Hermes is being operated by Claude Opus 4.6. You can see the context window, how full it is, but here it's just like any other chatbot or open claw you're talking to an AI model. You can say whatever you want. You don't have to use certain specific commands. Then to exit out of here back to the command line interface, you do control C. You might have to hit it once or twice, but this kind of it gets you out. So, now you're back here. Virtual environment Hermes root at the docker container and your folder is /opt/hermes. Let's type clear to get rid of some of that junk. Here we can type in Hermes space -h. This will give us the help file of all the things that we can do with Hermes. So, now what are we seeing here? These are commands that have to be typed in, you know, exactly as they are cuz now you're talking to the computer, not the AI model. These are flags or options can be things like a -h, --version, --yolo means that you bypass all dangerous command approval prompts. And some of basic examples that are commonly used. So, typing in Hermes starts the chat just like we saw. Hermes setup runs the setup wizard. Hermes model will run this in just a second. This allows you to select the default model. Hermes gateway allows you to run messaging gateways. Hermes update allows you to update to the latest version. Those are probably the most important ones. So, let's say we say Hermes model. It shows us this where we can select a provider and which model we want. News Research has their own. We are using open router. We can use Anthropic, OpenAI, Gwen, GitHub, Hugging Face, and more. So, let's say we select open router. These are the various models that they have on there. By default, we have Anthropic's Claude Opus 4.6. Very smart, very expensive. And you can go up and down and select whichever one you want. At the bottom, you also have a custom model names. And notice we have RCAI and Nvidia's NeMo Triton, which are free models. We're not going to get into local models here, but those are available. All right, so I'll scroll up back to Anthropic Claude Opus 4.6 the default because at this point it's like a dear old friend and let's hit enter. So because Hostinger already built out the Docker container for us, we already have some of our APIs in the system. For example, the Open Router and the Telegram token. Now, normally the first step would be to run Hermes setup. This would walk you through this installation wizard that would set all the settings and API keys etc. Technically, you don't have to do that on this install because the Hostinger Docker container already provides the bare minimum things they need to get started, namely Open Router API and the Telegram token. So you can actually continue to the next part if you wanted to, but I do feel like running in the Hermes setup and just going through it step-by-step might be a better way of approaching it, especially if you want to make sure that everything is set up right. If you're just trying to get to the working part as soon as possible, you can skip the step if you wanted to. So let's hit Hermes setup. So let's go through full setup. As you can see, we have our active provider Open Router and we can select the model that we want to use. So we're keeping our current model Anthropic Claude Opus 4.6. You can rotate between providers if one gets rate limited. Do you want to add an old credential for same provider fallback? This is a great feature. I don't need it right now, but I do plan to use it in the future. So I'm going to say no. Then we have our text-to-speech provider 11 Labs, OpenAI, MeMax, Mistral, and the Nutes. Currently, I'm using Edge Text-to-Speech and it's free cloud-based and no setup needed for the time being. I'm going to keep it. Might be fun to test out 11 Labs at some point in the future. They are really really high quality. So this is the terminal backend. Now since we're running Hermes in a Docker container, you'd think that we'd also want to run this in a Docker container. Local means it's running on the host machine. I'm going to keep it as is, but this is probably something to look into at some point. Next choice is the maximum tool calling iterations per conversation. Default is an idea which works for most tasks. 150 is deep research, multi-step coding projects, etc. Or as you see here, use 150 plus for open exploration. So this is where you want Hermes to just keep going. So 60 is probably a little bit conservative. I'm going to change this to 90 and if we're doing a lot of complex jobs, bump it up to 150 plus. This is a tool progress at display. Basically, do you want it telling you everything that it's doing? If I'm using Telegram and I select for example, all, it's going to mean that it's going to show every tool call with a short preview. I've been kind of liking it so far, so let me see how it goes and maybe I can change it later. And of course, the next step up is verbose. That's probably for debugging if something's going wrong. You really want to know everything that's happening. For now, I'll keep it at all and maybe turn it down to just showing new tools, but I'll hit enter here. Next is context compression. 0.5 is the default. Basically, compress it at the halfway point. Or do you want to compress it later? Go up to 0.95. I really get it right to the line. So for certain tasks where you really need a lot of the earlier details to be kept in the context memory, maybe it might make sense to go to 0.75 or maybe even 0.95, but that's going to cost more tokens and you risk the chance of running into the model's context limit. For now, 0.5 should be good. Next we have our session reset mode. So the default, which is inactivity plus daily reset, which sure comes first. That seems like a great idea. So basically gets reset at certain amount of time, at a certain interval of minutes or at the set fixed hour each day. Basically, you want to prevent the stale conversation history from bleeding over into a new tasks. So this seems good. I'll keep it as is. Next is the session reset policy. So messaging sessions such as Telegram accumulate context over time. Each message adds to the conversational history, which means a growing API costs. To manage this, sessions it can reset after a period of inactivity or at a fixed time each day. So the agent basically thinks through what's important, kind of saves it, jots it down, but then the conversation context is cleaned. You can also manually reset anytime by typing {slash} reset in chat. So this is important so you can always just hit {slash} reset if you remember to do that. This might be a good idea at certain points. So 14 40 minutes, that's 24 hours. That seems very long. I would think it would be in several hours and you would reset. If you have short focused sessions and then you on to the new thing, but I'll I'll start here and see how it goes. And I'll try to remember using {slash} reset when starting a new idea or or there's a new focus in the conversation. Daily reset hour, four, why not? Select platforms to configure. I have Telegram set up and I think really that's all I need. Reconfigure Telegram. If you've been having issues with it, you can reconfigure it. One interesting thing that I found is the Hostinger bot configuration. It doesn't put the Telegram bot token into the dot env file, which where usually things go. So for the time being I'm going to say no. So here it looks like we have two tools disabled on both of the CLI and Telegram. Let's see why and what those are. So for Telegram, a mixture of agents is disabled as well as reinforcement learning training. So this might be interesting for later discovery. You need an API key, that's probably news research that's backing this. I'm just going to hit enter for now and I think I'll leave this as is. So next you're able to type Hermes and start chatting or Hermes gateway to start the messaging gateway for your Telegram. Starting the gateway if it's not started is likely the next best move. So we're going to say no and Hermes gateway. So I'm just typing in Hermes gateway and hitting enter. Now it's already running for me, but if you haven't started yet, this might start it and then you're able to message your bot with Telegram. And now we can just type in Hermes to start the AI agent. And before you get too carried away, don't forget about the Botfather because he has not forgotten about you. Remember the Hermes agent we set up earlier? We got the API key for it, the bot token, and now we have to make sure that we're able to communicate with it through Telegram. Now, you can open up it as a new user and just type in its username through Telegram. So, I'll go to search or a contact or whatever and then type in the name you gave it, the full username, the one that has to end in bot. Or here from this menu from the Botfather, you can just click on this right here. Once you entered here, so this is the Hermes agent. Usually there's going to be a big button that says start. You click the start button and it wakes up the bot and you're able to talk to it. Or if it doesn't work, you just type {slash} start and that wakes it up. As soon as it wakes up, it's going to say, "I don't know you. Who are you? I don't trust you." Which is good because you don't want just anybody talking to your bot online. It's the internet. You don't want people talking to your bot. But it gives you this pairing code so they can pair your account and the bot so that it knows to trust you and to communicate with you. So, they're saying, "Run this command." Hermes pairing approve Telegram and then the code, the pairing code. So, you can just click on this to copy it to the clipboard, head back over to the CLI, the command line interface. So again, virtual environment on. You're in the Docker container. You can tell cuz it's a bunch of numbers and letters. It doesn't say SRV something. And then by default, you're in it {slash} opt {slash} Hermes. We type in the command Hermes pairing approve Telegram and then the pairing code. So, you can just paste it in there and hit enter. Now, I already paired this agent, that's why it's not finding the code. But for you, this would pair the two together and allow you to talk. If it doesn't, well, head to the troubleshooting section because we might have some solutions there. If any of this fails to work for some reason, we'll talk about it a little bit in the troubleshooting section. So, the first command that I gave the agent through Telegram was to find out everything about the environment, what hardware it's running on, what hosting, what skills. Give me a full rundown of your situation. And also, I told it that its name is Hermes, the emoji is whatever that face is. And I briefly described what kind of humor it enjoys to give it a little bit of a personality. I'm used to doing this from the open claw days. As you get more and more agents, it does help to give each one a certain quirk so you can tell them apart just by how they talk. Now, as it's running, sometimes it's going to ask you to approve certain things. You can click always, so approve always. You can click session, which is approve for this session, or just approve that one action. And of course, you can click the red button, which is deny or block or whatever it is, which basically means don't do that thing. So, our little agent figures out everything, who it is, what it's running on, the hardware, the setup, the virtualization, and hosting, that's on Ubuntu operating system, software, etc., etc., etc., and the 74 skills that it has. He says he's connected to me on Telegram, ready to go deep on whatever I need. The real question is, how hard do I want him to work for me? And whatever emoji that is. So, I got to say, I'm pretty sure he nailed it. Now, initially, it wasn't replying to my messages, so I actually opened up this command line interface to talk to it. Said hi and got initialized. And I just asked a question about its Telegram bot name. Now, my bot did run into some issues with using a Telegram right at the bat. When you install yours, you might not have that issue, you might have a different issue. Keep in mind that this is a bleeding edge tech that we're talking about here. Things break. If you have any issues, you know what I'm going to say, you paste it into your favorite chatbot. This is actually a really good time to talk about where all those credentials are saved, all those API keys, tokens, all this stuff that you don't want getting lost or uploaded somewhere, they go in your .env file. So, .env is just period and then e n v. Inside of it, you'll find probably something that looks like this, Anthropic API key equals and then the key, Telegram bot token equals and then the token, and whatever other stuff you put in there, for example, the Open Router API. So, we'll exit out of our bot with control C. You might have to hit it multiple times and we can run the command ls. ls is list. It will list all the files in that directory. Now, by default, the .env file will be hidden. Just so you don't accidentally mess with it. To see hidden files, you ls space -a. Again, your helpful AI assistant will walk you through how to do that and even tell you how to view it once you find it. If you don't find it, just ask and it'll actually give you code to run a search for it. By the way, if you ever find yourself sitting there and typing commands and trying to figure out how it works and why isn't it working, you might start wondering, is this normal? And the answer is yes. This is actually a surprisingly big part of the job. It's been made much, much easier with AI chatbots, but you you still have to handle it a little bit. Pretend you're a cowboy. So, I paste the command it gave me. It found that it was in the data folder. So, it's in opt/data and that's where the file is. So, you need to go up one directory and then go into data. So, type in cd space dot dot and as you can see, that takes us up a directory. We're now in /opt. If we do ls -a to see hidden files, as you can see, the data folder is there. So, we're going to do cd data and enter. ls -a. These are all the files inside of it and that's our .env file. This is the command that will show you what's inside that file. cat and then and of the file. Cat is short for, you'll never guess, concatenate. Concatenate is the correct answer. So, I'll paste that here. So, this is basically a very large file that contains all of your keys and various security credentials. Now, I got to say, New Research actually did a great job of just listing everything out. So, you're actually able to go through, add your Open Router Anthropic keys. You have your Hugging Face and various other providers that you can fill out. This is actually very useful. So, if you scroll down, they do have the Telegram integration as well as whatever else you want. WhatsApp, email, Gmail, etc. So, if your installation doesn't automatically fill all of this out, which again, usually it's will, but some percentage of us will have glitches, especially if it's right after a big update like what just happened. Well, then you're going to have to go in here and just add those keys yourself. So, the hashtag symbol means that this entire line is commented out. Commented out means that that code, that line, is ignored completely. Usually use that if you want to leave a note for another human that's reading the text, or if you want to use this code, but just not now, so you comment it out so it's there. If you just remove that symbol, then this becomes a working line with working code. The easiest way to change it is to use something called a nano. It's a very, very light word editor. So, it's kind of like notepad. So, you just say nano and then the location of which you want to edit. In this case, it's our .env file. All right, so, I'll paste that in here. So, notice nano space /opt/data/.env and we'll hit enter. Nano not found, since this is a brand fresh install of Ubuntu, they probably don't have some of these things installed, which is actually not bad. We quickly need to install it. Again, just type it into cloth. By the way, if you already have something like Open Claude installed on the computer, you can have it do all of this for you. I told Claude let's install it in Nano, it gave me the bash command to install Nano and updated. We just copy that and paste it here. And in just a few seconds, it's done. By the way, if you ran a command you want to reuse it, hitting up on the keyboard will cycle through the various commands that you have already ran. In this case, Nano we're using it on the .env file and we hit enter. Okay, it's working. It's glorious. So, we'll just hit down a couple of times until we get to here, then we'll hit delete a few times. So, notice it turns from green, which means it's commented out and ignored, into white, which means it's now going to be actual code. And we're going to put our API key here at the end. So, you just paste your API key there, and then we'll start scrolling down. All right, notice here we got to the Telegram integration. All right, first we're going to need our Telegram bot token. That's going to be the number here. It's in Telegram from the BotFather. So, just paste that in there. By the way, you can also put the allowed users here, so that'll be yourself or anybody else that's allowed to message the bot, as well as the home channel, the chat for cron delivery, so for those timed messages or those scheduled messages, and the name for the home channel. For the time being, I'll just add the bot token. All right, next we need to exit, so that's control X. And it's going to ask us, do we want to save? We push Y or N. In this case, I'll push Y for yes, and then it will show us kind of like, do you want to write to the same file that you're editing or create a different one? We're just going to hit enter. That's going to overwrite the existing file. Boom. We'll go back to Hermes and ask, can you see my Telegram messages? So, it knows the name that we're using for it. I'll ask if it can check that .env file to see if you have the bot token. So, it is able to see it, and good news, it actually doesn't just type it out in chat, which is terrific. If we didn't add the pairing code, we can just run in this approval for that pairing code. Control C to exit out of Hermes, and then we run just the command that it gave us. And it'll say approved user whatever on Telegram. By the way, once you update the .env file, I've learned the hard way you have to restart the Docker container. Hermes has said it can't do it. I'll have to see if that's actually true. But, if you run into that issue, just use this command of Docker Compose restart. So, after changing the .env file, run this command. And then, if you need to get back to the Docker and environment, you can ask again your favorite large language model. It gave me these two lines. The first one gets you into the container, the second one activates the Hermes environment. So, it's basically line one, and now we're in the Docker, and then it's line two. And that activates the virtual environment. Hit enter, and we're in. Now, I'm type Hermes, and we are good. I'm going to say see if Telegram works now. Every once in a while, this will pop up, and it'll say, "Do you want to run this dangerous command? Do you want to allow it once, allow for the session, or add to permanent allow list?" Well, I'll add it to a permanent allowed list for the time being. What's interesting is that little thing that pops up warning you. That's Hermes' little security bot. They call it Tirith, as in Minas Tirith. You don't know what Tirith means? Well, it's it's Elvish. It means guard. Did you learn something new today, or are you a nerd and you already knew that? Did you pass the dungeon wisdom check? Whatever the case is, it looks like Telegram is working. By the way, if that same guard message appears in Telegram, it will look like this. Allow once, allow for the session, or allow always, or deny. It'll list the problem here, and then you get to choose. So, let me know what you thought about that. Were you able to install it? If not, tell me what went wrong. I'm planning to do a full local installation guide soon, including how to migrate your Open Cloud install to Hermes Agent. What are you most interested in learning more about when it comes to running these Hermes agents and just whatever else you have on your mind. Please subscribe, hit thumbs up, and yeah, come back and see me cuz I feel like this is going to be big. I'm Miss Roth. Thank you so much for watching. I'll see you in the next one.