The EXACT System to Build AI Employees with Hermes Agent

[00:00:00]Hermes agent is one of the most powerful agentic orchestration layers on the internet right now. And if you use it properly, you have the ability to create multiple digital employees that work for you 24/7 around the clock and not only work for you 24/7, but get better over time. This is something that can learn from its mistakes, improve itself autonomously. It is the closest thing that you can create to a digital employee right now. In this video, I'm going to give you the exact five-step system that allows you to do that. And once you learn this, you'll have the confidence to build out any agent you want. The first part of this five-step system is to get your agent online 24/7 and basically set it up. So, if your agent cannot stay online 24/7, then it's not always going to be learning about you, learning about its workflows, or even be a reliable part of your business. If you want something that you can communicate to all the time, that's doing work for you all the time, then it needs to be online 24/7. So, you can build this locally, but it doesn't make much sense because Hermes agent is something that gets better over time, just as a full-time employee should get better over time, right? I'm going to show you why today we are setting this up on a VPS. So, first, let's compare the cost. Now, the cost is really the only benefit of using local. It is free and a VPS is going to run you5 to$10 dollars a month, but honestly the cost is trivial for what a VPS allows you to do. It allows for ultimate flexibility as you're going to see later down this list. So local is free, yes, but is free really worth it at the end of the day? I don't think so. Because if you want your agent to always be on locally, it's not always on. The minute you shut your laptop, the minute something happens to it, the minute you're away from your setup, your agent goes offline. on a VPS. It doesn't matter what setup you're using. You could be using a setup across the world and you could connect to your agent and communicate with it. And so a VPS is always on 24/7. Self-learning.

[00:01:57]Well, locally it is self-learning when again it's online and when it's open, but on a VPS it's always online. So yes, it's always learning about its workflows, improving on itself, and learning more about you and always ready to communicate with you and answer your questions when you set it up on a VPS.

[00:02:13]Now, let's look at the setup. Both of these require a one-time setup. So, you can set it up once and it will always be online ready for you to use specifically on the VPS. But locally, if you ever open up your laptop and you want to communicate, you can do that for sure.

[00:02:26]Now, let's take a look at flexibility.

[00:02:28]On local, flexibility is low. It's using up your processing power in the background. So, even if you do set this up locally and you want it always running, it's still taking up slots on your machine and it's always using computer processing power in the background. Well, on a virtual private server, you just put all that processing power in the cloud. So, your local machine stays very light, very agile, and you can spin up multiple different Hermes agents on a VPS a lot easier than you can on local. I left a link in the description to get to the place where you can access a VPS. What we're going to be using is Hostinger. And Hostinger is the sponsor of this video, which is very convenient because this is the exact service I would use whether they're sponsoring me or not. And I have Hermes Agents already set up on here.

[00:03:17]And they make it super easy for non-technical people to spin up a VPS or a virtual private server. So, what you can do is you can hit choose plan or you can just scroll down because it's going to take you down there anyway. And they have a couple of different options. This is what I love. You scale as you go. So you don't need to go to the KVM8 right away unless you get multiple multiple instances of Hermes agent. What I would recommend is a safe bet is KVM2. This is going to allow you to spin up multiple different agents. I'm going to be choosing the KVM2 plan because it's more than enough and you can do multiple agents. I'm going to select choose plan.

[00:03:53]For the period, I would recommend 24 months because this guarantees that your agent is going to stay online for two years and you pay once and you forget about it and you get the biggest discount. You save almost $400 when you go for that 24-month period. And you can pretty much uncheck all of the extra things if you don't want them. I wouldn't say that they're necessary. A lot of the stuff should be set up and ready to go for you. You can get an extra 10% off this annual plan by using my coupon code AI foundations. If you hit apply, you're going to get an extra 10% off, which is going to bring this to under $200. So now I'm going to hit continue. And then you can put in your payment information. And I will just hit submit. And then what it's going to do is it's going to get your services up and running for you. Since we're using Hostinger, they make it super easy.

[00:04:42]You're going to get this screen that says Hermes agent configuration. Make sure you save this somewhere. It could be in a password manager, a Google doc, whatever your security system is. We can get back to this, but we're going to need this later. So, make sure you save this information here. And then your VPS is going to take about 5 minutes to set up. But that's it. This is super easy, super handsoff. And basically what this is is infrastructure for your AI agent to run on so it doesn't go offline. Even if your laptop's off, you can still use your phone, which I'm going to show you how to set that up to communicate with this thing. And you don't have to be home. You don't have to be at your office. You can be on the go traveling, sleeping, and the Hermes agent employee is going to continue working for you.

[00:05:23]Okay, so my VPS has been created and now it's just deploying my Hermes agent and now it's running. That took actually less than five minutes. And just like that, you probably have the hardest part of the video complete. We have made sure that our agent can stay online 24/7. Now we need to move on to step two of this system. So now that we know that our AI agent is going to stay online 24/7, next we need to set up the communication layer, which is step two of this five-step system. So the communication layer involves a couple of things.

[00:05:53]First, we need to give our digital employee a brain. What do I mean by that? Well, we need to give it a model.

[00:05:59]And so you can select any model in the world that you want to power your Hermes agent. So that's step number one.

[00:06:05]Connect a model to it. So it has a brain. Step number two is our communication layer. So how do we talk to this agent? You can use Telegram, WhatsApp, Discord, uh I think WeChat, Slack. You can use whatever you want.

[00:06:19]I'm going to be teaching you how to set it up in Slack so you can have multiple different channels where your agent can post information too when it's doing certain things, right? You don't want your scripts for your content going in the same folder as your competitor research or your finance folder that has all of your financial data. You don't want to mix and match things. So, Slack allows you to create channels and your agent can use those channels to organize its thoughts and messages it's sending to you on a schedule or uh nondeterministically as well. And so, what you need to do is go to Hostinger Docker Manager projects. Your Hermes agent container is the one that we want to go to. And so, where it says access, don't hit open, but hit this button that says terminal. This is probably the most sound way to set this up. When you select terminal, it's going to spin up a web terminal on Hermes and open it in your directory. So, this is now your Hermes agent container that we are in the back end of essentially. And don't let this scare you. I'm going to walk you right through this to set this up.

[00:07:21]I'm going to just type in Hermes setup and then hit enter. And what that's going to do is bring up the setup wizard. And the option I'm going to select is blank slate because if you do full setup, it's going to run you through a bunch of questions and it's kind of confusing. So blank slate is the easiest and you can scale as you go with capabilities and what you need. So select blank slate. You can use your arrow keys to move around here. I'm going to select space or enter. First, we need to set up an inference provider.

[00:07:52]And so it's going to take you to this screen where you can input any model in the world. Pretty much what I'm going to be using is Open Router because Open Router aggregates all of the world's models into one platform. So I don't have to go to Anthropic to use Anthropic and then OpenAI to use OpenAI. I go to Open Router and I can change to Enthropic or OpenAI from there. And you have one key, one payment and it's super easy. So this is my Open Router account.

[00:08:19]It's openouter.ai is the URL. if you don't have it already. And what you can do is you can just load in $5 worth of credits and get started. I think you can load in literally $5. You don't have to load in 112. You can load in like five or 10 bucks and then get using any AI model you want in the world. They have over 400 models in here. And I'm not sponsored by Open Router. I just love using them. And so what I'm going to do is I'm going to create an API key by on the homepage selecting get API key and then selecting new key. And then you can name this key. I'm just going to name this Hermes video. Delete later, please.

[00:08:57]So, I do not forget to delete this. I'll set a credit limit of 30. So, this is like $30 worth. So, if my agent goes rogue or somebody gets my key and I forget to delete it, they can only spend $30 using this key. I will select create. And now, this is what's going to give us access or what's going to give Hermes agent access to any model in the world. So, I'm going to select copy. And then I'm going to go back to the terminal and select open router. And then it asks for that API key right after you select it. And so I'll paste it in. And then I will hit enter. And now I can select any model I want. And so here you can really see the pricing of the models too. As you can see, Fable 5 is super expensive and ChatGpt 5.5 Pro are super expensive. One of the best cheapest models right now is GLM 5.2. It has Opus 4.8 8 level quality with much cheaper cost. And so I'm going to select GLM 5.2 and hit enter. And then it says select terminal backend. What I'm going to do is just keep current. Hit local.

[00:10:00]And then it says your minimal agent is ready. What next? I can just select start with everything disabled. Finish now. And hit enter. And so now we can actually use Hermes agent. It won't be that good, but we can use it. So I can just in the prompt bar here type in Hermes and hit enter. And that will open up the Hermes agent CLI. As you can see, it looks very beautiful, very satisfying to work in here. But now I can just type in something like, "Hey there, what model are you?" And I can send it off.

[00:10:31]And the agent is going to initialize.

[00:10:33]And as you can see, it's online. We are communicating with Hermes agent. It says, "I'm running on GLM 5.2 via Open Router. I'm Hermes agent, an intelligent AI assistant. I help with coding, file editing, running commands, research, and a bunch of other stuff. And so, we've given Hermes agent a brain. And so, part one of setting up the communication layer is complete. Now, Hermes agent has the ability to communicate, but where do we communicate with it? So, when we send it a message, it can respond to us. So, let me show you how to set up Slack. You can do this on Telegram. I have a video in the upper right hand corner if you do want to watch that where I go over how to set this up on something like Telegram. But in this video, I'm going to use Slack. And so for this Slack connection, I'm going to walk you through the real Slack setup on the Hermes agent documentation. And so you can use this if you ever get stuck as well. That'll be a nice resource to have. I'll leave this down in the description. But step one is we need to create the Slack app. And we can do all of that manually. But what Hermes does is it makes it super easy to generate one of these by just copying this line from option A, going to your terminal, and pasting it in. And so that's what I'm going to do. I'm going to copy step number one from option A in the documentation and then I'm going to go back and what we want to do is we want to open up that terminal. Just if you were using the terminal like we did in the last step where we connected the model, you can close out of there and just open up a new one because what we need to do is we need to get right here.

[00:11:57]Now, I will say a lot of the time this won't work if you're using the legacy version of Slack, which I actually am.

[00:12:02]I'm not using agent view. And so if you're not using agent view, you need to get rid of that line. And so that's what I'm going to do. I'm going to get rid of the agent view line. If you just paste this in there and you get some error and it doesn't generate the manifest, then you'll understand, okay, I need to get rid of that agent view line and try again. And so what I'm going to do is I'm going to paste that in here. I'm going to use my arrow keys to go back and I'm going to delete agent view. And now I just have Hermes Slack manifest-.

[00:12:31]And then I can hit enter to send that off. And what that's going to do is actually create the manifest. And as you can see, it says Slack manifest written to. And then it gives me the exact file path where it was written to. And then it gives us some next steps for what to do with this file. Now, right now, we can't even see the file. So, before we go into the next steps, we need to make sure we can see that. In order to do that, you can just copy this little line right here. Try not to add any extra spaces or anything. Just copy this line directly. I'm going to hit control C or command C on Mac. And then I'm going to type in c a t cat and then space. And then I'm going to paste that in. So cat and then paste that in. And then I'm going to hit enter. And what that's going to do is it's going to print the entire manifest copy and pasteable right into your terminal window. So this is like the JSON manifest that it created right here. And this is what we need to paste into our Slack app in order to prepare it for Hermes agent. and from this bottom curly bracket all the way up to the top curly bracket. Copy all of that to your clipboard and then open up claude in the separate window or chat gbt and then you can just prompt it something like this. I just received my slack manifest and it's a JSON object.

[00:13:44]Is there any way you can format it properly in a code block so I can easily copy and paste it to my clipboard. Then I can paste in all of that uh manifest right here and it might have copied okay but I just want to make sure. So I'm going to send that off. You don't need to be on high. You can be on low for that. Send it off. And the reason I'm going so in depth on this is because this is a super important part of the process. What we're doing is we're creating the communication layer and allowing us to communicate with any employee that we create in the future.

[00:14:11]This is kind of like a one-time thing.

[00:14:12]You can use multiple Hermes agents and multiple Hermes profiles out of this app right here, which is really cool. And so now it's putting this in a nice copy and pasteable code block just to make sure that it's exactly the way that Slack likes it. I could not imagine writing all of this on my own. I am blessed to be in an era of AI. Okay, so it's done.

[00:14:32]Once it's finished completely, make sure it finishes completely. Hit that copy button in the upper right hand corner of the code block. And now you have your Slack manifest ready to go. And so then we can just go back and we can follow the instructions from the Hermes agent documentation. So it says the next step is to go to Slack and create an app. So, I'll select create app. And make sure you click that link in the documentation because it's not slack.com. It's api.slack.comapps.

[00:15:00]And if you're not logged in, it'll have you log in and create an account. You can do this on a free Hermes or uh rather a free Slack account. Um, but then you can hit create app. And instead of doing it from scratch, we now have that manifest that's going to create it for us. So, I'm going to select from a manifest. I'll pick one of my workspaces. Hit next. And then this is where you want to paste in that manifest. So I'm going to hit command or control A and delete everything. And now where it says paste your manifest code here, I'm going to paste in that entire file which is 395 lines. It's a lot here. But then I'm going to hit next.

[00:15:35]And then we uh have our OOTH scopes all set up and ready to go. And everything's all good because we did that manifest.

[00:15:42]So I'll hit create. And now you get your app credentials. You get your app level tokens. You have a name. You can even add an app icon right here, which I like doing. So, I'll do that real quick. Give your assistant or your employee a face.

[00:15:54]I'm going to choose a picture of me with glasses on, and it's a cartoon image.

[00:15:58]So, this is my Hermes agent now. Pretty cool. And I can even change the background color to match my shirt a little bit more. Maybe make it like that nice dark green. But yeah, once you have uh customized your Hermes agent to your liking, you can hit save changes. And now the beautiful thing about doing step one in the way that we did it is we get to skip steps two through five and we can go straight to step number six. So after you've edited your agent to your liking, you can go to step number six where we install the app to the workspace. And this is very simple. It's going to tell you to install it to a workspace. you're going to get a bot user ooth token and then you're going to copy that token because that is our Slackbot that we just created. And so I'm going to go back to our settings page right here in our Hermes agent where we were editing it. Then I'm going to go to install app in the sidebar over on the lefth hand side. And then we're going to install to AI Foundations. And so this is my workspace which I've already signed into. I'm going to hit allow. It's basically like you're signing into your account or giving your agent access to your account here. And then you get a token. Anytime you make a change to your Hermes agent, if you want to change a scope, give it more permissions or less permissions, you will have to reinstall and do this OOTH all over again, just like we did, which was very simple. But you can copy this token and put it somewhere. After you've saved that token somewhere, you need to find your user ID for the allow list. So this is very easy. You can just click on the user's name or avatar, view their full profile, click more, and then copy member ID. So here I am in my Slack app.

[00:17:29]To find your member ID, just go to your profile in the bottom lefthand corner and then go to profile. Again, we have this lovely picture of me looking all confused outside, but where you're going to find your member ID is in these three little dots right here and then underneath my head, probably in the bottom right hand corner where you see the blue, there's a button that says copy member ID. What you want to do is copy that and then save it somewhere.

[00:17:52]I'm just saving my member ID right there in my OOTH token in the other tab. Step number seven is done. Next, we need to configure Hermes agent. So, there's a couple of ways to do this. If you're just using like local, I would recommend doing it this way. It's a little easier, but the way we're doing it where we're on a VPS, I would recommend the Hermes gateway setup. It's going to be the easiest way to just get prompted on when to put your tokens in and it's going to be super nice. It'll store those credentials for you safely. And so, what I'm going to do is that. Now, I'm going to go back to this terminal where we did the manifest. And actually, I'm just going to open up a new terminal. If you ever want to do like a new command, I just recommend opening up a new terminal so you don't have to navigate your way around and you can have a fresh terminal. So, I'm going to go back to Docker Manager projects and I'm going to go to access terminal. And what we're doing right now is we've generated the Slack app. We've installed Hermes using the manifest to our Slack account. So, we now have an agent sitting in our account, but this is all backend stuff.

[00:18:50]Now, we need to get the actual Hermes agent gateway set up. The gateway is the link between the manifest and Hermes agent in our Slack. Everything is connecting allin-one with this Hermes gateway setup. And so I'm going to type in Hermes gateway setup. Another thing you could do is you can just go copy Hermes gateway setup right here. But it's not too hard to type in. So that's what I'm going to do. Hermes gateway setup. And then I'm going to hit enter.

[00:19:18]And then it's going to prompt us with what platform do we want to configure.

[00:19:22]And if you don't see Slack initially, just scroll all the way to the top because it's probably hiding up there.

[00:19:27]So, as you can see, Slack is second from the bottom on my page, but it could change for you. And so, then once you're hovered over Slack, you want to hit enter. And then all of those things that we need are going to prompt us to type in. And so, we need a Slack bot token.

[00:19:43]That was the OOTH token that I told you to copy and save somewhere. So, mine's right here. Temporarily saved it in my URL. I'm going to copy that. go back and paste it in. If you do not remember where it's at, you can go back to your Hermes agent Slack app that you created and go into install app and then you have your bot user ooth token right there. That's okay, too. But once you paste it in, you want to hit enter. And then you need your Slack app token. And the Slack app token is something we have not done yet, but it's very easy to set up. So go back to your Slack API where you set up your Hermes agent and then scroll down on the basic information page to app level tokens and hit generate token and scopes. You can name this token whatever you want. Call it Hermes backend app or something. Doesn't really matter but add scope and then what you want is connections. Right?

[00:20:36]This is pretty much the only thing you need activated for the scope is connections colon write. Then you can hit generate and now you have your token. You can hit copy, done, and then go paste that into your web terminal.

[00:20:49]I'll paste that in and then hit enter.

[00:20:51]Now we can add our allowed user IDs. The reason we got our user ID is because this is the person who's going to be able to message Hermes agent. So I if you want another person's account or a team member to be able to contact your Hermes agent, make sure you add their user ID. And you can always change this later as well. So I'm going to copy my user ID that we got from my profile and paste it in. And that's not a secret. so it's not redacted and I can just hit enter. And now we have home channel ID.

[00:21:19]We don't have to set a home channel ID, but every Slack channel has a specific ID attached to it. And so if I were to create a channel in here and call it like Hermes Research, I'll make the visibility public by going to the upper right hand corner, going channel details, open channel details, and the ID is down here. So, if you wanted this to be the default channel that Hermes responded to, you could do that very easily just by putting in a channel ID that's actually in your Slack. So, Slack has these channels where different conversations or threads happen. Another thing you can do is you can go to your DMs right over here on the left and then you can just type in your bot name and hit enter and then you can open up this.

[00:22:02]You can message this real quick.

[00:22:04]Something like, "Hey," and this is a channel ID. So, what I could do is go to the upper right hand corner of my Hermes bot and then I could hover over app details and then open app details and he has a channel ID right here. Yeah, I'm calling him a he. That's all right. So, I can copy that channel ID and have my DM with Hermes be the default thread he responds to as well, which is what I'm going to do. And so, I'm going to go back to Chrome. You don't have to do this either. You can set that later and you can just invite your bot to channels to respond to. But I'm going to set my DM channel with my Hermes app to the home the home channel ID where all my responses go. I will hit enter. And now I will hit done. We are set up on Slack.

[00:22:44]And so now we need to run that gateway so we can actually communicate. What I'm going to do is open up one more fresh terminal. And we're almost done with the hardest part of this video, but I'm going to open up this terminal again.

[00:22:54]Start fresh. Now I'm just going to type in Hermes. Hit enter. And then once we're within Hermes, we can actually type in this command to our AI agent and that is slashgateway.

[00:23:08]And what that's going to do is give us a command that we need to start the gateway. So I'll hit enter. And as you can see, it says once we're within Hermes agent, by the way, it's hard to find that without actually going to Hermes agent, but it says to start the gateway, you can do this right here. So, I'm going to copy this Python command.

[00:23:28]Python CLI.py-gateway.

[00:23:31]Going to copy that. And then I'm going to X out of the Hermes chat. You do not want to paste this in the Hermes chat, but you want to be out of the CLI. And so, I'm going to hit control C. And that says goodbye. And so now I'm out. You can again go open a fresh terminal if you're having problems getting out of that CLI. But then I'm going to paste that in and hit enter. And what that's going to do is enable the gateway. What I can do is I can say, "Hey there, who are you?" And Hermes agent should reply with, "Hey, I'm Hermes and I am using the GLM2 model or something like that.

[00:24:09]Hey Drake, I'm Hermes agent, your AI assistant. I can help with a wide range of tasks." So now the gateway is running and it's pretty much ready to go and we have our AI agent that we can communicate with here in Slack. If I wanted to invite it to a channel, I could even do that. I can type in slashinvite. Then I can add agents and apps to this channel. I'll select that.

[00:24:29]And then I'm going to go down to my Hermes agent and I will select add. And so now Hermes joined this channel. And anytime I want to communicate with Hermes, I can at Hermes and then say something like, "What's the weather in Michigan?" Then I can send that off. And when I at Hermes agent, it's going to be able to respond, which is very, very cool.

[00:24:50]And so very quickly it replied. I can view the thread. Now we can see Hermes agent working. It's extracting the weather and generating a response. And so we're getting a lot of smoke over here in Michigan right now. So a lot of smoke is in the air from those wildfires in Canada. And then it gives me a little summary here. So this is very cool. And so the thing about that step that we just went through, it's a very difficult and quite technical step, but I want you to think about it like this. Now you have a communication layer and you can create multiple Hermes profiles in multiple Hermes agents all in one workspace and they can be replying to different channels. They can be replying and talking amongst each other. So we've basically given Hermes agent, one of the most powerful harnesses in the world, a place where it can communicate with you and other AI models. And that's step two. Things are going to start picking up speed here now. So step number three of this five-step full-time employee system is giving Hermes agent a role.

[00:25:46]So, I wouldn't recommend having one agent that can do everything like script writing, coding, research, personal tasks, etc. I would recommend breaking up your Hermes profiles into specific roles that do one thing very well, that have one sphere that it learns very well. So, I wouldn't have a Hermes blogging agent also be uh analyzing my YouTube video stats. I would want like a YouTube analyst, an Instagram analyst, a Tik Tok analyst, all gated off in their sections, getting better over time with their domain that they're best at. So, the next part is kind of going to happen in your mind. What roles do you want Hermes agent to fill? What tasks do you want it completing for you and getting better at on a daily basis? Because this is like a 30 to 60 day process and even longer of Hermes constantly learning about the way that you like to work and get things done. And so that's step number three is deciding a role for this agent. This is when things truly start to get fun and you can start building out your workforce. So in order to add a Hermes profile, what you need to do is you need to go back to your Docker manager, go to your projects, and then you need to select this open button. Now you're going to need your credentials I told you to save earlier. If you didn't save those, don't worry. Just select manage. And you can get to the screen from here as well. And then scroll down to environment. And then you're going to have an admin username and an admin password with the values to the right of it. What you need to do is you need to hit this little link. Same place as the open button on the Docker Manager projects. You can hit that. Either one takes you to the same spot. You want to click that and this is going to take you to a Hermes agent backend which actually is very nice to work out of. If you never have signed in, it's going to take you to a screen that looks like this.

[00:27:32]you can just paste in your username and then you can go again down to your environment and grab your password in that same way. So now I'm going to sign in and what that does is it takes you again to this backend where you can actually chat with Hermes agent. I can say hey there and what it's going to do is actually respond to me which is pretty cool. So you can communicate with Hermes in here using the CLI. You can look at all of your past sessions that you've had with Hermes. So anytime you've asked it a question you can look at all of your files which is very cool.

[00:28:01]You can select models and you can even configure different model settings like what models do you want for auxiliary tasks mixture of agents and other things. You can set up cron jobs in here which are scheduled jobs for Hermes to complete. So this backend dashboard is super nice but what we need to focus on and yes we need to get focused is on the profiles. And so in the sidebar you're going to see a tab that says profiles. I want you to click on that because these are your individual Hermes agents and they're called profiles. So, we have our default one. That's just the agent that we set up in the beginning of this video. Before I show you how to create a new profile, let me just show you what is involved with a profile. What does a role look like in Hermes? Well, every profile has a specific model that it uses. If you have a coding agent, you're going to want to use a heftier model that's good at software development. If you have an agent that's just focused on tracking your habits and your personal productivity, you don't need that big of a model because it's very basic task it's completing. So model matters. You get to change your model in the profile.

[00:29:07]You get to edit the description. Then you have your soul.md. This is the system prompt or personality powering that specific profile. So I have a very basic one right now. You are Hermes agent, an intelligent AI assistant. You can change the soul MD in your Hermes agent as well. you can manage your skills and tools. And so, we're going to get into those. Those are the next steps of the system. But first, it's just about these three things up top. What model do you need? What is the description? And what is the personality? And this sets the direction for what you need. So, what are you needing most in your workforce right now? Do you need a writer? Do you need a researcher? Do you need a coding agent?

[00:29:46]Do you need all three? This is where you want to decide. What are the roles that you need? And whenever you decide there's a certain role that you like or a certain role that you want, what you can do is you can hit create in the upper right hand corner when you're on the profiles tab. And then you can give it a name. It gives you instructions for how to name your agents here. Let's just say I want a research agent. What you can do is you can clone your config from any one of your profiles that you already have. I like doing that because a config is kind of what we set up with the Slack thing and uh any keys that you might have that you want to share across agents. That's going to be important.

[00:30:25]So, I'm just going to clone config from default. I can give a description. I can say this profile is really good at researching and specifically competitor research. So, if I ever want to do competitor research, I'm going to make this guy the agent that can do it really well. And then it says model. So, we get to select our model. And we already handled the description here, which is kind of cool. The model we can inherit from clone or we can just use the open router key that we have in our account already. I'm just going to say for the research assistant, maybe I want to use a Grock model. Grock 4.5. That's a great model. And then I will hit create. Now, as you can see, we have two profiles in here. If I hit commandR and I refresh this page or I just refresh in general, what you're going to notice is in the upper lefthand corner now we can switch between the default agent or the research agent and we can view the research agents sessions and we can view the default agents sessions. And so each one has its own sessions, its own tasks that it's working on. Now we've already set up the model for this research agent. We've already set up the description. Now, we need to think about a soul MD. What is a good personality for a research agent?

[00:31:43]I'm going to go back to my Hermes agent and ask it. I'm going to say, "Hey there, Hermes. I'm creating profiles right now manually in the back end. You don't have to worry about doing it for me. What I want to understand is what would a good soulm look like for a research assistant that is going to be good at researching my competitors?"

[00:32:03]You can ask Hermes because Hermes has knowledge on Hermes and you're going to see that real quick. So, when I send that off, as you can see, it's already looking into this. It's doing approval commands on the documentation. And then it gives me a competitive research assistant instruction file, which is very cool. We're creating agents with agents now. It says, "A strong soul.md looks like stable, broadly applicable, and specific in voice, not overloaded with temporary instructions." And so it gives identity style avoid in defaults.

[00:32:32]And instructions are meant to be quite small. They're not meant to be very long in the soul.md file. It's meant to be very direct. So I'm going to copy this.

[00:32:40]I'm going to go back and then I'm going to try pasting this in to see if it will work. Okay, it looks like it is going to be fine. If I hit save, now we have created a research agent with a personality. And we might want to adapt the personality to our exact workflow.

[00:32:55]But the point is is I'm trying to show you this research agent is locked in on research.

[00:33:01]This default agent that we created in the beginning of this video that has a bit of a different job and you can go about it in multiple different ways, but I'm going to show you the way I like to think about my Hermes profiles. This is the idea that I keep in mind as I'm building out profiles in Hermes. I think about any command that I might want to give to an agent or any question I might have. This command is always going to be going to my general agent. My general agent is the one you just saw me message in Slack that created the soul file that did basic stuff for me like told me how to write a soulm and all of my requests.

[00:33:42]I don't want to have to go searching for my agents. I want all of my requests to go through this agent and then have this agent's job be to route my request. So any command that comes to this general agent, he can pick the profile that gets the job done best and send the question to that agent and then get me the answer from that agent. So these are like subworkers under one chief of staff.

[00:34:07]This is like a chief of staff executive assistant that knows how to route my commands to the right department. And so each one of these could be a Hermes profile. If I have a financial question, I ask my general agent and then he asks the financial agent. And this kind of agent routing system is really what makes a good workforce. And so that's profile creation in a nutshell. Now, this is an important step because it really trains your mind to think differently. Instead of having one AI agent that accomplishes every small task for you, you create these sub agents that can get better over time and have their own environment to work in with their own tools, their own skills, and everything in between. Which leads me to step number four of this five-step system that is really going to help this process become much more smooth. So, a profile without these next two things is nothing. It really doesn't matter. you can create the best system instructions.

[00:35:03]The only thing that's really going to change is the personality unless you add steps four and five to this. And so step number four of this five-step system is to add tools to your agents and specifically your profiles that you create because each profile is going to need a different set of tools and each tool provides a new capability to your agent. So, here I am in my Hermes backend just as we were before. And what I want to do is I want to switch to this research agent because all of these sidebar options are going to change as I switch my agent in the upper leftand corner. And if you do not see the profiles that you create, make sure that you refresh the page. When you refresh the page, it will load them in. And so, what I can do is I can switch to this research agent. And after I switch, I'm going to hit refresh anyway. And now, as you can see in the upper leftand corner, it says managing profile, research, config, keys, skills, MCPS, model, and new chats apply to that profile. And so, I'm in my research agent right now. And I'm just going to walk you through what I might want to do. I'm going to go to skills and then I'm going to switch to tool sets. And tool sets is where all the fun happens. As you can see, browser automation completely inactive. Web search and scraping completely inactive.

[00:36:16]We have file operations which is good and terminal and processes which is good but vision and image analysis inactive.

[00:36:23]So there's a lot of things we can add in here. Some of which aren't free. Some of which require an API key. But there's a lot of things like memory that we should just have active so it can remember things about our previous session. And so I'm going to hit configure on memory.

[00:36:38]Memory is a really good first tool to add. And I'm just going to go through here and think about what are all the things I want this research agent to be able to do. Memory is really important.

[00:36:47]So I'm going to configure that. And this is easy. All you have to do is flip that switch. And now memory is enabled for that agent. Another really good useful tool set to activate is cron jobs. Cron jobs allow you or allow your agent to create scheduled tasks. So maybe it wants to set up a competitor research task every morning at 7 a.m. where it scrapes all my competitors, gets their videos, their transcripts, yada yada. I can hit configure on crown job and enable that. And now that agent can create scheduled tasks for me on my behalf. Session search is another one for searching past conversations I've had with the agent. Going to turn on session search. Task planning is important because task planning allows the agent to plan out step by step before it actually executes, which is really good in benchmarks for multi-shot prompting and reasoning. So, I'm going to enable task planning. I don't need image generation, video generation. I don't even really need vision and image analysis, but web search and scraping would be nice to have. So, I am going to hit configure on that. And it looks like we need to set some things up. We could use things that aren't free like Firecrawl. Firecrawl is a really good web scraping tool but it's not free. We could use things like Tavali which is free to start but then paid. We could use XAI's web search but that's also paid. Or we can just find a free tool that does it for us. And so what I'm going to do is duck.go DDGS.

[00:38:14]And it says this backend needs a one-time install. Runs on this host. May take a few minutes. So, I don't really know whether to do run setup or select first. I'll just run setup and then it will install and then I think I'm going to select it. Okay, so it looks like it finished running the setup. Now I'm going to select it. And now make sure you scroll up and you enable web search and scraping. So now we should be able to search the web, which is very cool.

[00:38:39]And so this is how to add just general tool sets that Hermes agent will give you out of the box. And you want to do this on both of your agents. That's why it helps setting up your default agent from the start with all of these active and ready to go because when you create your profiles, you can clone your profile config and all of the tools to the next profile so you don't have to keep on doing this over and over and over again. You now have connected some basic tool sets and that can get very extensive. You can connect your Gmail, your Google calendar, and pretty much any tool in the universe that you want.

[00:39:09]Just ask Hermes agent how to connect it and it will walk you through that perfectly. The next thing that you need in the last step of this five-step system are skills. And skills can be created. And again, Hermes has a bunch of pre-installed skills. A skill is an automation that your agent can run in order to complete a task in the exact way that you need it completed every single time. And this is what it's going to autoimprove. That's what makes Hermes agent so good is it creates automations for you to get things done in your style and in your taste and then it will improve on that. So if you give it feedback and you say I don't like that, it will inject that next time it tries running the skill and improve it to make it better to get you a result that you actually like. And so this is the final step of this five-step system. Let's get into actually creating a skill and looking at all of the skills Hermes has to offer. So skills, just like tools, can be configured for any AI agent that you have. So, if you have a research agent as you saw me set up, if you have a scriptor agent, a default dashboard agent, whatever it might be, you have the ability to create skills for that or to activate a lot of skills. Now, you usually have around 60 plus skills active to begin with. And you can scroll through here on the skills page and look at what all of them do. And you can even ask Hermes, hey, what does the blog watcher skill do? What does the claw design skill do? And so what I'm going to do from start to finish is I'm going to connect a tool, which is going to be my YouTube analytics tool. I'm going to tell it to look for competitors based on my channel. And then I'm going to tell it to create a competitor research automation that watches my competitors and sends me a message to Slack uh every morning. And so I'm on Vid IQ. Vid IQ is a tool I've been using for over four years now in order to help me grow on YouTube. And I'm not sponsored by them.

[00:41:01]I just love using them. And what I'm going to do is create an API key. So, I'm going to hit generate new key. And I'm just going to call this uh competitor research. And then I'll put delete so I can delete that later. And then I will copy my key. And then what I'm going to do to add this tool to my Hermes dashboard is I'm going to go to keys.

[00:41:25]And at the very bottom of keys, you have a spot where you can add custom keys.

[00:41:29]So, this is going to take you out of Hermes agents base tool set and it's going to uh create a custom key for you in your profile. And you do want to make sure you are on the right profile. So, I'm going to move off of this dashboard, my default agent, and I'm going to go to my research agent. Now, when I scroll down, I want to add my custom key. And so I'm going to name this vid iq_appi_key and then hit add. And now I can enter a value. So I'm going to paste in my API key. This gives it access to my account.

[00:42:07]And if you're going to connect Google calendar, that that one's a little bit more difficult, but if you're going to connect a tool like notion, air table, the process is going to be very similar for a lot of tools where you have an API key in the back end that you need to connect. And so I'm going to hit save.

[00:42:23]And now my Vid IQ API key is set. And now what I'm going to do is I'm going to tell Hermes to set this up for me. And so I'm going to tell it, hey, I want to connect to Vid IQ. I have put my API key in the config. Can we set this up for my research agent profile? You can say this is an MCP server and I need you to connect the config.

[00:43:00]And so anytime you want to connect a tool, you not only can have Hermes help you through it, but you can tell it to connect the tool for you, which is very, very cool. And so I'm going to copy this vidiq.com/mcp the documentation. And I'm going to also go give it back to my Hermes agent. And then I'm going to send it off. And just like that, it says uh done. It's added the MCP server to my research profiles config. So now I have a YouTube tool that can do YouTube research for me, which is great. And it knows my channel since I've connected my channel to it.

[00:43:33]What I'm going to do now that my Vid IQ API key is connected and it says what's confirmed working. Research profile loads the correct model. Vid IQ MCP server connects. Everything is all good and ready to go. So, what I want to do is I want to set up a skill, an automation that can scrape my competitors. But first, I just want Hermes to ask questions to my research agent. What it's doing right now is in the back end, it's running a Hermes-P research. And so, -p actually spins up my research profile. This is being used in a terminal, and it's a command that you can usually run in your terminal on a computer, but since we're on a VPS, we can run the terminal in the cloud. And then it says chat QQ your query here.

[00:44:16]Beautiful. Then it loads in the profile soul model config and everything it has.

[00:44:21]What are my five latest videos? And I don't really have to say anything else because it's my Vid IQ account. So it knows about my channel already. And my default agent is not doing anything other than sending the command to my other Hermes agent. And it says here are your five latest videos from AI Foundation's channel. I didn't give it any context about me. Remember, I am just using the tool that I've connected to my channel. And then it gives me my five latest videos with all of the views, subs, likes, comments, everything. I can ask Hermes, what I want you to do is find my top 10 competitors. And then I can send that off. And so it gives me what it thinks my top 10 competitors are right now.

[00:45:02]They used Vid IQ, similar channels, neural similarity, ranking, and confirm stats. So, Vid IQ is an amazing tool and it gave me these people as my AI competitors. Regardless, I'm going to say, "What I want you to do is create a skill that takes the top five of these competitors and watches their channel every morning, and it gives me a competitor research uh briefing every morning, and I want you to post it in the Hermes research channel. And I want my researcher agent profile to be doing this so it can get better over time because that's the whole point of Hermes agent is to have your agent get better over time at what it's doing. So I don't want this default agent to be doing all of the work and all of the learning. I want that agent too. And so now I'm going to send that off. And look what I'm doing. I'm saying I want you to create a skill. A skill is a step-by-step way you go about things in order to get a certain result. Now Hermes is just going to work on everything it needs. competitor channel IDs, Slack channel info, and research profiles skill directory. It's very easy to create things like this. So, it ran through, it did its work for quite a while, which is fine. And then it said what I built. Everything is built, the skill, the crown job. And there's only one thing left that I need to do. And so, the skill is located within my profile. So, I can actually go look at that skill within that Hermes dashboard I was showing you earlier. And the skill instructs the research agent to fetch all of the stats for my competitors and use my Vid IQ MCP tools. Analyze signals, format everything in Slack Markdown, include an insight of the day, etc. And then it says, I need to set up the research profiles gateway if I want that to run on a schedule. But look at this last line. It says, "The skill is designed to get better over time. As the research agent runs more briefings, it'll learn your competitor's patterns, refine which signals matter, and improve the inside of the day quality. I don't think I have to set up this gateway either. What I can do is I can say, can't you just set up a cron job and trigger the prompt to run at a certain time every day with the profile agent in the query? Because in similar fashion how it contacted the research profile to begin with this Hermes agent can send off a message at a scheduled time to that agent and it says you're absolutely right. That's a much cleaner approach.

[00:47:17]The default profiles gateway is already running and can fire cron jobs and that's this agent's only job is to trigger prompts. The research agent itself is going to understand okay I post my message to here when I'm finished. It said fire it off. The crown job ran successfully and is being delivered. And now I have a notification in Hermes research which I can look at.

[00:47:36]And now I'm in my Hermes Research and this is the competitor brief that it sent me. Daily competitor briefing versus my top five competitors. My channel 350K subs, got my views, videos, everything I need. And then it has a table which we could get formatted better. I know it uh doesn't look that great, but we can get this looking really cool and even add buttons so that maybe we want to make a video script from uh the signals that it's providing here. But it gives the top five channels, what they've been doing, their new uploads since the last briefing, and who's actually uploading. So, very cool.

[00:48:09]And I can make this look better, like I said, and I [clears throat] can even update the output that it's giving me.

[00:48:13]But the cool thing is is this skill is going to improve on its signals. So, if I wanted to, I could say, I like this, I don't like this, change this, change that, and it would auto update that skill for me, and it would have a better briefing next time. So, this has been the process of creating a full-time digital employee using Hermes agent. If you follow these five steps every single time where you are getting your agents online, you only have to do that once.

[00:48:37]That's with Hostinger. And then you get a communication layer, which you also only have to do once. Creating profiles, connecting tools, and creating skills.

[00:48:45]This is the format that is going to lead to success. And Hermes agent, I'm telling you, it gets better over time.

[00:48:51]Don't expect the first results to be amazing or life-changing, but work with it. Tell it what you like. Tell it what you don't like. And over time, it's going to shape into something that you actually like and use every single day, something that actually is helping you run your business. So, I hope you enjoyed. If you did, please drop a like and subscribe. Check out Hostinger.

[00:49:10]They're the best platform to host a VPS on. And with that being said, I'll see you in the next