Hermes Agent - Ultimate Crash Course for Beginners (AI Agent)

[00:00:00]This is Hermes, an AI agent you can run locally or in the cloud that's meant to have superior memory and self-improvement capabilities to things like OpenClaw, which I find crashes all the time. It's created by Nuance Research, a group out of America that have published it open source under the MIT license in February of 2026, and it's been wildly adopted online and for good reason. We're going to have a look at how to install it, set it up, and what makes it so special. Let's begin.

[00:00:28]This is the Hermes website. I want to copy over this command over here in order to install it. There is a desktop app and I'll show you that later. But for now, let's open up the terminal.

[00:00:37]From here, I'm going to rightclick on the terminal, select paste, and enter in that command. Then hit enter to install.

[00:00:44]And that's the installation done. Now, let's take a look at the three ways to set Hermes up. Quick, full, and blank. I want to do the full setup because if you select quick, you'll need to set up an account with Hermes on a plus, super, or ultra plan. And even if you select the free plan, you're going to need to attach some credit card details and pay as you go, which can end up being a lot more expensive than connecting it to a platform like Codex from Chat GPT. So here in the configuration, let's go ahead and select full setup. We're going to select our own provider. We have options here like from LM Studio to run a local model or from Anthropic or even Open AAI. As a quick note, we're not going to use any models here from Claude because currently these will not work with Hermes or even OpenClaw as Anthropic has put new restrictions on third party apps which require you to actually use the API rather than your regular subscription. So instead, we're going to use Codeex to set this up. Open AAI's codeex is pretty generous when it comes to usage. And on top of that, there is an additional step or two to do during the setup for security, which is great. The first step will be to head over to this authorize link, which will require you to log in. But be aware, you might get this prompt here that you need to enable codeex authorization for devices. To do this, you'll need to just jump back into your regular chat GPT subscription. Go to settings, and then from here, you'll want to go to security and login. Scroll to the bottom and enable the enable device code authorization for Codex. And this will allow you to then enroll your device back in that same URL we had previously to authorize it. It's going to ask you for the authorization code which we had here in our terminal. So just copy that across. And once you've done this, it's going to link the device to your Hermes setup. And now you should be able to use it inside of the app. Here we'll select a model to use. I'm going to use a GPT 5.5. And then we'll select how we want to run Hermes, whether it's locally in a Docker container or something different.

[00:02:44]I'm going to just run it locally, which is the default setup. Now, to configure the messaging, this is how you'll interact with Hermes. And you can do this in a few different ways, such as setting up WhatsApp or Slack, but my personal preference is Telegram since it has a easy configuration that we can set up. Now, you can do this manually by contacting BotFather, or you can copy out a QR code or URL. So, I'm going to do option one, which is to grab that URL. Here is that URL. I'm going to open it up here on the right hand side, which takes us to the link to create the bot.

[00:03:19]If you have WhatsApp running on your phone or on your computer, you get to this screen over here where you can give your bot a name. I'm just going to give it the name Hermes Agent. And this has now completed that setup. You can see here on the right hand side that we can now allow it to use this telegram account as the main bot to interact with. And we'll also confirm that it can only interact with the user ID which is myself. I'll select yes to kick off the gateway as a service so it's always running. And with that done we've set up Telegram. We have a couple of other things we can configure here but most of these are preconfigured so we don't have to touch these and we can just select done. Now after this, select yes to run the Hermes agent as a gateway service in the background and also reload your shell with this command over here which you can copy paste right back in. With that done, we should now be able to interact both with Telegram as well as run the command Hermes inside of this shell here. So let's try that out here on the left hand side for this Telegram message. I'm just going to do a hello world to see if I get a reply. You can see here at the top the bot is typing.

[00:04:28]It's running on GPT 5.5 on codeex and I've got a reply here to my message, hello Adrian. Now I know that not all people use telegram. Some people like to use these agents directly in the terminal or through a web UI. So let's have a look at how that works. Over here we can simply type in the word Hermes and this launches the terminal version inside of this terminal window. Here you've got a little chat interface and I'll do another test with hello world and we've got our reply here. Hello, how can I help you? The terminal might be also a little bit complicated to always interact with and I personally prefer the web UI. So here I'm going to ask how to start the web dashboard and the command for that is just to run Hermes space dashboard. Let's actually exit out of this by hitting Ctrl + C and then pasting in this command Hermes dashboard. This starts building the web UI which then launches here in the background as you can see in my browser and we can now interact with Hermes directly through this which is a lot easier to do so simply click on chat on the left menu and here we've got another version of that terminal but this also doubles as our chat interface so we can start working with Hermes right in here now let's take a look at this dashboard on the right hand side here we can create new chat sessions and this allows us to see those sessions on the right hand side as history as well as create them. The chats themselves look like they're running inside of a terminal because that's more or less what's happening here. On the right hand side, we can also change the model that we're currently running. Here we've got OpenAI codeex and we can choose which model under that we have. So, for example, I'll swap here to chat GPT 5.4 and you can see that's updated now with the new version. On the right side here, we can also change reasoning. I normally like to have this on high, but if you want immediate responses, you can turn it off. Or you can even set it to a setting above high, but let's leave it on high for now so it can actually think in the background when we send it messages.

[00:06:22]When doing chats, there's quite a few different commands you can utilize. I'm not going to go through all of them, but if you want to take a quick look, you can always type in for/help and get to see what they are and what they do. They include everything from starting new sessions, clearing old ones, saving history or compressing it as well as starting up agents and a lot more even like updating the app itself. Now, the first main change I want to make is for the interface itself. If you have a look at the far bottom left, then there's an option here for the Hermes style. Right now, the theme itself is this green theme, but I much prefer the blue theme using just regular system sands as the text is a lot more readable and useful for me to view it in this more light version. However, that's more a personal taste and for this video, I think the regular default Hermes teal should suffice since it's just a little bit more readable for this video. Next up, I want to go through the lefth hand side menu because there's quite a few different options that you can start using inside of Hermes. So we'll go through each one of these options to get a better understanding of how they work.

[00:07:24]First is sessions. Sessions are the chats that you have with Hermes stored with history. And each one can be independent such as this one from Telegram or this one we had inside of the terminal. Chats are also saved here into history. So you can review them in the future, download them, or even reference them in future chats or delete them if you don't want that information to be saved. So, if I head back over to chats and create a brand new chat and maybe engage with it, send a couple of messages and head back to sessions, you'll see that it gets saved in my sessions. You can view it here called testing sessions. I can actually view those messages, but for example, this one's not very useful. So, what I'm going to do is simply delete it. And that way, it's no longer saved in memory for the future. For the most part, you won't need to do much here because Hermes manages most of this itself, but it's just useful to know that it's there. Next up is files. Hermes does have its own directory and it's something that you can open up inside of VS Code or for example cursor if you want to manage that yourself. Hermes is stored under Hermes/ slash. Most of the things that happen inside of this directory can include things like saving skills, updating skills or chron jobs, and just configuration files as well.

[00:08:43]For the most part, you're not really going to need to manage much of this.

[00:08:47]And Hermes does a pretty good job at self-managing all of this. The next thing is models. And this is the models that we're currently utilizing inside of Hermes. You can see that OpenAI codeex is currently selected. But if you wanted to select another default model and select switch, it would restart the service and use that model. Auxiliary tasks are also based on models that you can select. Most of these default to the main model, but for example, if you want images or voice to be generated or using a different model, you can select that.

[00:09:18]There is also agents, which allows us to configure what models are used for agents that Hermes might spin up of its own accord. But for the most part again, most of these things can be left as default. Later in the video, I'll show you how to set up a local model like from LM Studio or Olama and connected to Hermes. The next tab is for logs. This again isn't probably something you'll need to access very much at all, but it is useful for the Hermes agent itself to have access to its own logs so it can identify what's going right, what's going wrong, and make its own updates.

[00:09:50]After that, we have cron. probably one of the most useful tabs on Hermes and something you'll probably want to check out quite often if you're using it. It allows you to create scheduled jobs. You can ask your Hermes agent to do this on your behalf or you can do it manually yourself like I'm doing right now. You can create scheduled jobs for pretty much anything from every minute to every hour or day or week or month. And here's one that I'm creating right now. Every day to run at 9:00 a.m. which is going to give me a morning brief. Right now, this is a pretty simple job, but I can work through this to do things like check all my emails, check my calendar, and maybe even check things like Slack and just give me a brief of everything I should be aware of for the day ahead, which can be incredibly useful when you set it up quite well. But for now, let's delete this schedule job and move on to the next tab, which is skills. Skills are basic instruction sets, like a text document, telling Hermes how to perform a certain action. You can turn these on or off and you've got a few predefined ones like this one here for claw design which is currently turned on and I've just turned that off but we can turn that back on and take a quick look at it. It's essentially all text but it's essentially like a mini harness or instruction set of how it should go about creating a document like should it use Tailwind CSS or Bootstrap should it make the file responsive or not and lots of other stuff. You can create your own skills. For example, if you've got a company or a certain and you've got a very strict way that you want the agent to operate a certain piece of software or an API. And you can also learn skills from online directories. Never before has documentation been as important as it is now. Like for example, this website called type UI has a different types of web designs you can generate based on a skill. The skill itself expresses how the UI should work in terms of spacing, coloring, topography, and much more. But just based on that, we can create a visual design that looks coherent throughout a different types of prompts or requests just based on utilizing that skill itself. As part of skills, we also have tool sets. These are things that agent can do such as scrape the web, use the browser to perform different things as well as the terminal or the file and folder permissions to say read a file, delete a file, or write to a file. All of these are tool sets that you can enable or disable depending on the security implications you want to allow Hermes to have access to. After that, we have plugins. I won't go over that too much, but essentially it's things that are baked into Hermes to utilize, such as memory and context here. These two are automatically built in, and they will compress the chat context window so it doesn't get too large, as well as the memory as well. Next, I want to set up some custom tools for the AI agent to have access to. MCP tools is how you do this. MCP tools are pretty important if you want to connect your agent up to third-party apps or services you use and also if you want to define the permissions it has access to in those services. So let's actually set one up right now so you can understand how it works. To do this, head to the top right here where it says add server. And here we're going to give it a label. I'm going to be adding a Zapurin cuz it's actually a great service to have especially if you're going to connect up to third party apps. and you can just get it here at zapier.com. Let's actually head over here into the section for MCP. There is SDK which is a little bit newer and I'll show you how to set that up a little bit later on. But under MCP, just select get started here on Zapier MCP. Let's create a brand new MCP server. There are some pre-existing one for things like clawed code and openclaw, but we're going to do one here under Hermes which will just be categorized as other. The MCP server is more or less created. Now, we can add apps and tools to it. And there's a wide variety that you can connect from Slack to Notion to your calendar. I'm going to connect Gmail as an example. What makes a Zapier special is that you can give refined permissions to what it has access to, such as finding emails and maybe creating drafts, but not sending emails or deleting emails. I don't want to come back to my inbox and find it all deleted. And it's one of the reasons I like Zapier because of these fine grain controls. I could go ahead and add more apps and tools in here, but for the time being, we'll use this as an example.

[00:14:05]Next up, let's head over here to connect. And this is the part where I'm going to connect it up to Hermes. I'll select to generate a token. And here I've got instructions on how to connect it up to Hermes. Now, I'm going to use the URL with token just cuz that's a little bit easier, but you can set it up however you wish. Heading back here to Hermes, I can simply paste in this URL, select add, and we're done. Then selecting the Thunderbolt here, it actually connects up the MCP server, grabs the tools it can access, and sends them over to the Hermes agent so it knows about them. These are the tools are right here. Let's give these a test.

[00:14:40]I'm going to head over here to chat in our regular Hermes chat for the agent.

[00:14:45]I'm going to ask it to grab my latest email and see if it automatically starts using the tools. In the log here, you can see the tool call to call the Zapier Gmail find email tool. And here is my latest email from MongoDB telling me how I recently logged into it. MCP does allow you to connect up quite a lot of tools, but it can get complex if you start adding in hundreds of different servers. And I personally just prefer to connect one which has everything already there. While MCP is useful, we're moving slowly towards SDKs. And there is an SDK you can use for Zapier 2. MCP is usually just for chat, but SDKs are better for agents because they're faster, more responsive, and more expansive on what you can do. So, let's actually set one up here for Zapier instead and just see the comparison. I'm going to copy this prompt and paste it here into the terminal. And it's actually going to load up the skill to set this up. It's going to run in the background. And once complete, we're ready to go.

[00:15:39]Realistically, the SDK is just like a terminal command that you can run now to call the tool. So, I'm going to ask it here to grab my latest email. And it's going to be able to do that using the SDK much faster. And here is our response. Moving on, let's take a look at the next few options here on the left panel. We have channels, which are essentially the messaging channels that you can talk to Hermes on. We've configured Telegram, but there are other options here like Discord, Slack, and WhatsApp. I'm not going to go through to configure every single one, but it's a pretty simple process, and you can sometimes just ask it to configure itself inside of the terminal. There are web hooks which allow you to actually connect up subscriptions to things like Telegram and Slack, but I'm not going to do that right now either. Next, we have profiles, which is essentially the configuration that you currently have set for Hermes as a baseline. You can add a description as well as a solem, which is essentially how Hermes believes itself to exist, whether it has a personality, some quirks. Normally I leave this as default, but you can configure this to take on a personality of a character or person you like. You can also change the model and other things here, but these are just the basic profile for it. If you want to get more advanced, you have to go to the config. Here are all the settings under the hood. Things you'll probably never need to configure unless you're a senior developer really looking to customize this service. But you can configure everything from how many characters it might use in requests to how long it'll wait for a timeout to how many agents it'll spin up and everything in between.

[00:17:10]But for the most part, I would leave this unchanged. Moving on, we have keys, which are where our secure tokens and keys are saved when you're needing to do requests like with OOTH or if you're connecting to services like GitHub or other services in general. After that, we have system, which is just what your current configuration of your hardware is. Right now, I'm just running on my Mac OS. I can also see that there's an update here, which you'll normally see on the bottom left. Now, let's actually kick that update off so that we have the latest version of Hermes running.

[00:17:42]That'll run in the background. So, let's continue going through this side panel and hopefully we'll be running our latest version of Hermes very soon. And finally, that takes us over here to documentation, which gives us a link to the actual website and its documentation over here. As you can see, right now, the documentation says that there's a new desktop agent that we can utilize.

[00:18:03]So, let me go on to the main website and install it and show you. It's actually really quick and easy to install. Just download it, drag it into your application folder, and then run it.

[00:18:14]From here, find it in your applications or just quick search for it. When you launch it, it'll open up this Hermes agent installer, which actually goes through the same installation process that more or less happens inside of the terminal. The benefit of this though is that it comes with an application window, which is a lot easier to manage than a web dashboard or just working inside of a terminal itself. Let's launch that so I can show you what I mean. On first launch, it's going to ask you to install a couple of dependencies such as Python and Node, but it'll do all of this for you in the background, which also is a nice little touch. Once complete, you're taken to this dashboard over here. And it comes with a lot of benefits, such as having its own little chat dialogue rather than working through something that looks more like a terminal. Lots of UI elements like on the right hand side there's a dashboard, there's settings, which has a guey interface now. a lot easier to customize and view in this way. And it feels a lot more organized, too. Previously, Hermes was essentially like a terminal with maybe a web UI added on. And this actually turns it into an application similar to what we see with things like chat GBT and Claude now having their own apps too. Feel free to explore and have a look at how this works. But overall, most of the settings as well as the configurations here are the same as the web UI. But rather than using terminology like channels, we're using things like messaging, which makes a lot more sense. Now, I want to take a look at how to set up Hermes in the cloud, which is actually pretty easy, but also useful, especially if you don't have a spare machine or if you're worried about security on your own machine, having an AI agent access everything. So, let's take a look at how to set this up. I'll be using Hostinger for this. It's the same thing I use for OpenCore and a couple of other services and I even run my own startups on it. It's got some of the best pricing around. I've added a link to it in the description below. And if you want as well to do it manually, you can use the coupon code aden Hermes which will give you 10% off the total price and it's really only about $9 per month which is extremely cheap especially for a VPS. What's useful is that it comes with a pre-installed version of Hermes, which means that we're not going to have to go through all that setup we did previously. It's just going to be ready and installed on the VPS straight away. So, let's actually go through this checkout and start the installation. With a VPS, you'll want to select a location that's closest to the country you're in. I'm in Australia, so I'm going to select this one here, which is in Indonesia. Then, I'm just going to go and select to do the Hermes installation as an application. You can just search it up here at the very top. But if you want to manually install it on YUbuntu or something, you can do that, too. It's just going to be a little bit more complicated. I'm going to select Hermes Agent here, and it'll come with everything that I need, including the web UI. I'll be able to log this a little bit later on. This next step here is probably the most important. Make sure you set a username and password that you'll remember here. I've got the pretty default that I have, and I'm going to copy this into a document to use. Make sure you don't leak this online or on a YouTube video because this will give anyone access to your Hermes setup and they'll be able to log into it at any point. This step here will set up in the background for a couple of minutes and finally take you to this dashboard here where you can click over here to access your Hermes agent. For this step, just sign in using the username and password we had earlier on. And now we're back into Hermes just like we were when we installed it locally. From here, you can go ahead and start chatting to it or for example, connecting it up to your own API keys or even setting up a couple of channels to interact with it through Telegram or WhatsApp. Because Hermes runs on the local workstation, this is probably one of the most secure ways you can set it up. And that way, one day, if you do want to get rid of Hermes or upgrade or use something different, you can simply literally just delete this virtual private server and you'll never have to worry about it again. Now, we're just scratching the surface of what you can do with Hermes. There's lots of use cases you can apply and I've done a full video of use cases just up here which you can check out. It is on open claw, but all of them will carry across the same because this is essentially just a different harness.