Hermes Agent Fundamentals In 29 Minutes

Welcome to the Hermes Agent's Fundamentals Guide. Setting up your

Fundamentals Guide. Setting up your Hermes Agent correctly is so important.

It's the difference between having a terrible time versus being able to build something sustainable that you can build on top of and become more magical every day. It's been over 4 months now that

day. It's been over 4 months now that I've been using Hermes Agent consistently. Here's my multi-agent

consistently. Here's my multi-agent system doing research and building workflows and softwares while I'm chatting with it using local models to keep things free and private. Not that I have anything to hide, of course. And

here is my Obsidian second brain and getting smarter every day. So, in this video, I'm going to walk you through my Hermes setup. I have tips and tricks

Hermes setup. I have tips and tricks along the way that can make a really big difference for your experience. And I'll

be giving you exact prompts and workflows so you can also customize and build your Hermes Agent to exactly what you want it to be based on your budget, requirements, and use cases. Now,

without further ado, let's go. All

right, outline for today's video. We're

going to start off with talking about hardware choices, where your Hermes Agent is going to live. Then, I'll walk you through the download and setup process. We'll walk through Hermes major

process. We'll walk through Hermes major features like tools, skills, and cron jobs, and the memory management system including how to optimize it. I'll also

talk about use open-source models. And

finally, after you've mastered these fundamentals, I want to give you a taste for the multi-agent frameworks, how powerful those can be, especially in combination with OpenClaw or Claw code, which is absolutely a game-changer if you're building software. By the way, I

just want to say thank you guys so much for suggesting me moving over from OpenClaw to Hermes. I honestly was and still am a big fan of OpenClaw and will probably revisit it again later. But for

now, Hermes has become my major local agent. So, yeah, you guys were right.

agent. So, yeah, you guys were right.

Okay, let's start off with talking about hardware.

For me, I primarily run my Hermes Agent on my Mac Studio. It has 64 gigs of RAM, an M4 chip, and it's on 24/7. This is

powerful enough for me to run some pretty great local AI models and only accessing more powerful cloud-based models when I need to do something like really fancy. This is a good balance for

really fancy. This is a good balance for me and my team because even though the Mac Studio did cost me a lot of money, it is still way cheaper than when I was using a Mac Mini or like even my laptop and having to rely really heavily on

cloud models. But do not fear if you

cloud models. But do not fear if you don't happen to have a Mac Studio. You

can also have very powerful Hermes Agent setups with other options. I would say that there are four primary options that you can choose. The first one is a dedicated powerful local machine like a Mac Studio, for example. The second one

is a VPS, a virtual private server, which is kind of like renting computer space on the cloud. These can be as cheap as just five or six dollars and be on 24/7 as well. The third option is just to use any old computer or laptop

that you happen to have around, like this MacBook Pro, for example. 16 gigs

of RAM. This is actually where I started. Just make sure that you wipe

started. Just make sure that you wipe everything and I just kept it like on 24/7, so it was basically like a dedicated machine. And finally, if you

dedicated machine. And finally, if you really don't want to spend any money and you don't happen to have any other computers lying around, you can also use your own personal computer. But, if

you're going to do this, I highly, highly recommend that you use Docker, which containerizes your agent, so it doesn't allow it to go bananas on the rest of your computer. All right, so I'm not going to go into way too much detail about each of these options, uh because

I have actually covered those in some videos, which I can link in the description you can check out. Now,

let's actually move on to the next step, which is setting up our Hermes Agent.

[music] So, to set up Hermes Agent, you want to go to hermesagent.newsresearch.com

and you have two different options. You

can install it via the desktop app if you have a Mac or you can install it via the terminal. You can have a Mac as well

the terminal. You can have a Mac as well or if you have a Windows. Most tutorials

I see actually go through the terminal route. So, I actually want to show you

route. So, I actually want to show you how to do it via the desktop app, which is a great option for people who don't feel comfortable typing stuff into terminal. But, if you are going through

terminal. But, if you are going through the terminal, it's super easy as well.

Just click the copy, stick it into your terminal, and just go through the onboarding sequence. So, click download

onboarding sequence. So, click download for macOS. All right, so once you have

for macOS. All right, so once you have it downloaded open, it would look like this. A clean Hermes Agent. This

this. A clean Hermes Agent. This

interface probably looks very familiar, it's very similar to your normal chatbot interfaces. And the first thing you want

interfaces. And the first thing you want to do is come over here and look at the different models. So, there are some

different models. So, there are some models that are already here because I have like an Anthropic key that's set up. This probably going to look

up. This probably going to look different depending on what you already do have set up over here. So, you can just choose what you have and if you don't have anything that you like, you can also choose edit models and then

click add provider. Here you can connect to any providers that you have. If you

have an open AI ChatGPT subscription, you can just log in to that and be able to use your subscription here or any of the other ones here. If you want to go fully local, that also works. I will

show you a little bit later in the video how to set up your local model. But, if

you don't happen to have like any kind of provider subscription, the easiest option is probably just go with the news portal. It is a monthly subscription

portal. It is a monthly subscription that gets you access to hundreds of different models. Very convenient. And

different models. Very convenient. And

once you have that connected, you can just start off by saying, "Hi."

Yay! Just had your first interaction with Hermes agent. By the way, the general consensus is that the Sonnet 5 model is the best driver for the Hermes agent. But, because of Anthropic not

agent. But, because of Anthropic not letting you use your subscription and you actually have to use the API key, it can get pretty expensive. For me

personally, because I use Hermes a lot, my main driver is an open source model called the Quant 5.6 35B model, which I run using my Mac Studio. There's a lot I can talk about when it comes to model

selection and what you can be using, but to keep ourselves on track, I'm just going to put on screen now some of the common models that people choose to use as their main drivers, including their pros and cons. And I'm also going to put

in the free guide below more detailed recommendations as well as a prompt that can help you choose what is the best driver model for your specific setup.

Now, let's continue with talking with the agent for a bit. The first thing that I actually recommend that you do is tell your agent a little bit about yourself. Now, instead of actually

yourself. Now, instead of actually typing it because of AI psychosis and I'm incapable of like writing stuff anymore, what I like to do is go to a chatbot like Claude, for example, and

then just write like, "Give me a brief description about what you know about me that I can give to Hermes agent." So, yes, it will just

tell me about myself, which I can just copy that and I'll write, "Here is a bit about myself. Please remember this.

about myself. Please remember this.

There you go. And you can see that it says running memory because it's actually going to remember this. Now, you may be asking, "How is it

this. Now, you may be asking, "How is it remembering it? Where is it storing this

remembering it? Where is it storing this information?" Great question. Let me

information?" Great question. Let me

show you. I'm going to actually going to jump now to my Mac Studio, which has my like actual like proper Hermez Agent running up. I I, you know, redownloaded

running up. I I, you know, redownloaded and installed it again to show you guys like the download process.

Uh but let's actually see what it looks like. I'm going to move there now. Now,

like. I'm going to move there now. Now,

this [clears throat] is the Mac Studio.

I don't know if you can actually notice a difference, but anyways, let me actually show you where Hermez Agent stores his memories. So, if

you go to finder, if you have a MacBook, it is stored under the users, your name, so my username is Tina, and then .hermez. And you can literally see

.hermez. And you can literally see everything that's contained here. But

this is not very pretty, so I'm going to actually show you using Obsidian. So, it

looks better. So, here is the .hermez

folder, and it will be telling information about the user, me. It's

stored under memory, and you can click user. So, this is literally like human

user. So, this is literally like human readable. You can read it yourself what

readable. You can read it yourself what Hermez knows about you. It has specific information that it saves into the .users. It's just a markdown file. And

.users. It's just a markdown file. And

since we're talking about where it's storing stuff, there's also a memory markdown file where it saves information that is very crucial to your current setup. This is the built-in Hermez

setup. This is the built-in Hermez memory, so you don't have to do anything here. So, I will talk a little bit more

here. So, I will talk a little bit more about the memory system and how it is that you can enhance this a little bit later in the video. By the way, there are a lot of things that you can do just via this desktop app. It is so convenient. Like if you go to your

convenient. Like if you go to your settings over here, there's a lot of things that you can configure. So, your

model, information about your chat, you can assign it different personalities, appearance. Uh there's a lot of things

appearance. Uh there's a lot of things that you can customize, what you want it to look like. I just like leaving it as default. I think it looks pretty good.

default. I think it looks pretty good.

Information about workspace, safety, memory, and context. Yeah, there's a lot that you can actually do just from here.

And with that, let's transition into the next section. I want to show you the

next section. I want to show you the major Hermes features so that you can have a very strong foundation to build on.

Okay, so there are five major features that I think is important for you to understand. The first one is tools. So,

understand. The first one is tools. So,

tools are pre-configured infrastructure that's managed by News Research, the creators of Hermes. You'll be able to do things like search the web, generate images, text images, browser automations, etc. So, it's not something

that you actually do yourself. For

example, I can ask Hermes agent, "Search the web for research into the latest AI chip developments."

chip developments." So, we see that it uses a search function and was able to get the results on the latest chip developments. If you

click on skills and tools and you look at tool kits, here you can see a lot of different tools that Hermes natively has. So, this is already very, very

has. So, this is already very, very powerful, but Hermes can do more than this because it also has what are called integrations. Integrations allow Hermes

integrations. Integrations allow Hermes to call so call upon and use other third-party softwares. It's what allows

third-party softwares. It's what allows you to connect to different providers, for example, for different models, like News Portal, Anthropic, OpenAI, MiniMax, etc., etc. It also allows you to connect

to messaging platforms. Like over here, I have Discord connected. Ignore my

1,000 tabs, but this is how I like to get alerts and communicate with Hermes as well. So, here Hermes is sending me

as well. So, here Hermes is sending me my daily brief.

Good morning, Tina. Here's your AI brief for Thursday, July 4th, 2026.

That's a little taste for what you can build using these integrations. It's

also where I get alerts for bills that are successful from Hermes, too. By the

way, if you also want to connect to Discord or like Telegram or any other chat app, it's really easy. You can

actually start a new session and just ask your Hermes agent, "Hey, can you help me connect to Telegram?" For

example, or WhatsApp, or whatever.

And if you click on messaging, you can see all the different types of message integrations, messaging app integrations that you can connect to if you wish. I

personally like using Discord uh because it allows me to be able to separate like different channels and then get like alerts and communicate based on different channels. Also, I can use it

different channels. Also, I can use it on my phone. But, I'm pretty sure that they're probably going to roll out with a mobile version of the desktop app soon as well because Open Cloud did. So, I

wouldn't be surprised if you can just use the desktop app like with a mobile version very shortly too. Great. So,

News Research and Hermes supports a lot of these integrations already and they're super easy to use. I'm going to put on screen now some of my favorite integrations that I like to use. And if

it so happens that Hermes doesn't have a native integration already, you can use what is called MCP in order to provide your agent with additional third-party tools. Notebook LM, for example, is one

tools. Notebook LM, for example, is one of my favorite MCPs in order to do research and to create a podcast like you just saw earlier for my daily briefings. And the way you get it to

briefings. And the way you get it to connect is also, as you probably would guess, really easy. You can just say, "Hey, help me connect to the Notebook LM MCP, [clears throat] for example." And

then you would just follow the instructions. I'm also going to put on

instructions. I'm also going to put on screen now some of my favorite MCPs and what I like to do with them. I'm not

going to go into too much more details about MCPs and any additional plugins. I

feel like that's a little bit more advanced, maybe for a next video. Now,

let's move on to the second category of features. Also, the feature that made

features. Also, the feature that made Hermes famous to begin with, which is skills. You see, Hermes agent advertises

skills. You see, Hermes agent advertises itself as the agent that grows with you.

And what that means is that it's able to learn as you work with it, aka develop skills with you over time. So, what are skills? Well, you can think about skills

skills? Well, you can think about skills as an instruction manual for completing a task. If we go to Skills and Tools,

a task. If we go to Skills and Tools, you can actually see some of the pre-built skills that Hermes already has. And there's like different

has. And there's like different categories for this. For example,

there's Apple Notes that can manage Apple Notes, Apple Reminders, iMessage that can send and receive iMessages.

Let's see, if we go to Research, there's one that allows you to search different papers, Creative lets you generate ASCII video videos, etc. etc. There are a lot of them over here. So, you can actually check them out. But, what makes Hermes

famous is that as you work along with it, it's able to learn these skills both by itself and you can specifically teach it new skills to do, too. Here, for

example, in this conversation, I was asking Hermes to help me research some business ideas and see how valid it is.

And it tells me this is ideal for creating AI-generated micro-dramas, which is something that's already very dominant in production in China. So, it

you know, did a lot of research, told me information about what's the landscape looking like, what the competitors are looking like, what's the blueprint, and where's the opportunity here. And it

also gave me a plan for what to build, including phase one, phase two, and phase three, blah blah blah, etc. I then asked you to give it a final score after scores across different dimensions based upon these additional criteria. So, I

gave it more criteria about what I cared about. Uh so, I don't want to raise VC

about. Uh so, I don't want to raise VC capital, I want it to be bootstrap lifestyle business, etc. etc. And then finally, it was able to evaluate how good this is for a business idea that I might want to pursue. So, it tells me that it has a score of 7.5 out of 10,

and more details over here as well.

Here's where I want to draw your attention to. I asked it, "Make this

attention to. I asked it, "Make this into a skill for evaluating business ideas." And that's when I said, "Sure."

ideas." And that's when I said, "Sure."

And it was able to make this skill. You

can see the description of the skill.

So, what it's supposed to do. So,

business idea evaluator is supposed to score the framework for evaluating business ideas against bootstrap lifestyle criteria and the different dimensions. And it also explains how

dimensions. And it also explains how Hermes in the future should be using the skill. So, how to use. User describes a

skill. So, how to use. User describes a business idea. Um it would score each

business idea. Um it would score each dimension 1 to 10, assign weights, calculate weighted total, and then present the total score out of 10, individual scores, and key strengths and weaknesses, and then finally give it a clear recommendation, which is go,

caution, or no go. And so now I have that skill available to me, and I can do {slash} it's called business idea evaluator, and I can say evaluate business idea. Uh let's see. To evaluate

business idea. Uh let's see. To evaluate

business idea of a personalized learning app that is able to generate a study plan and lessons specific to what you

want to learn that uses AI to be able to generate this. Okay, cool. And we can

generate this. Okay, cool. And we can see that the it says here the user has invoked a business idea evaluator skill.

Okay, great. It tells me lots of things, blah blah blah blah blah, and it tells me, "Wow, 9.2 out of 10." Great. I

should do this. And it asked me, "Want me to start building the phase one MVP?"

So, say I really like this idea and I'm going to use another skill and say, "Save this research about the AI learning app into my Obsidian vault."

Great. So, it uses the Obsidian skill.

If you want to look under thinking, it says, "I'll use the write file tool as recommended by the Obsidian skill." And

under projects, we have another project called AI learning app where it has a lot of information here. Don't worry so much about this Obsidian setup. I'm

actually show you later how to make this in the memory section of the video. But

first, I want to talk about one more major feature of Hermes agent, which is cron jobs or scheduled task. A cron job is a scheduled task that runs automatically at certain times or

specific intervals without having to manually trigger it. You see a lot of things in life are actually reoccurring.

Like every day, you got to get up and like exist and eat food, you know? Yeah,

like computer, when we're doing AI stuff, it's very similar. You generally

want to do very similar things every single day or every single week where you know some type of cadence. That's

why cron jobs are really useful for that. One example of this is a daily

that. One example of this is a daily scheduled brief that I have. At 11:00

p.m. the previous day, it creates this brief for me of all the daily news that I'm interested in and also makes it into an audio file that I can listen to in the morning. I also have cron jobs that

the morning. I also have cron jobs that summarize my Apple notes because I really like like, you know, writing a bunch of stuff into my Apple notes um and then compile that together into Obsidian as well. So, I can like kind of query my own brain if that makes sense.

Another one that's very important is that I have a cron job that every single night looks through to make sure that everything is running properly cuz I'm hosting a lot of the models. I also have like multi-agent systems going on

building software all at the same time.

So, I want to make sure that everything is being handled the way that's supposed to be handled. I will put on screen some of the cron jobs that I have and the ones that I'm building as well, some of my favorite ones, which hopefully can give you inspo for things that you can build as well. So, take a screenshot or

not. I also do have a free guide in the

not. I also do have a free guide in the description where you can sign up and then I will send you the free video companion to your email, if you want.

Great. So, we've covered the major features for Hermes Agent and you can build so much just with these features here. Now, there is one more concept

here. Now, there is one more concept that I have separated into its own section because it is so important, which I'm going to talk about now, which is memory.

For those of you who have used Open Claw or like Co-work or other types of agents previously, you probably got really frustrated at some point because of the memory. Like it started like forgetting

memory. Like it started like forgetting things. Um it can't like search up what

things. Um it can't like search up what it's supposed to start to getting really slow as well and these things started breaking as well. Very annoying. All of

this has to do with memory management.

Now, Hermes already has a pretty good memory system and actually like for most people, if you just use his default memory system, you would be okay. But,

if you are someone who really wants to get a lot out of Hermes and just use it to the best of its abilities, there are also two customizations that you can do, which I find that has been amazing and really, really helpful. But, first let

me actually quickly cover what is Hermes Agent's default memory setup. Okay, we

are back in Obsidian where I'm showing you the entire Hermes folder, everything that's contained in here. The memory

system is a two-tier system. And

actually, you've already seen uh the first tier when I showed you earlier.

Pop quiz, write in the comments, do you know what it is? It's okay if you don't.

I'll show you again. It is the combination of three different files.

The memory.md file stores really important information about the setup, like where things live, what are the major plugins and integrations that are there, what's in progress, etc. The

user.md file that has information about

user.md file that has information about the user, you, and the is sold on the file, which is information about the agent. So, it knows who it is. So, this

agent. So, it knows who it is. So, this

first here gives a general overview for your Hermes agent to know high-level information. But, each of these files is

information. But, each of these files is actually kept deliberately brief.

Otherwise, they just get really, really bloated and slows down Hermes in the long run. But, what happens if you want

long run. But, what happens if you want to find information that is more specific? Well, luckily, Hermes has a

specific? Well, luckily, Hermes has a second tier memory system where it can use a tool. Remember what a tool is?

Called session search where it can search through logs of archived information. This is stored in a sequel

information. This is stored in a sequel light database. I don't have to go into

light database. I don't have to go into too much detail about this, but Hermes actually does store all the conversations that you have about it. It

just wouldn't actually go and look at it unless it needs to retrieve the information. Let me actually show you

information. Let me actually show you this in action. So, start a new session and I will write, "Remind me, what were we talking about earlier about business

ideas?" And you can see that it's using

ideas?" And you can see that it's using the session search tool, searching past sessions. And it's able to tell you,

sessions. And it's able to tell you, "Okay, we talked about two business ideas recently, a vertical micro drama condensation engine and AI personalized learning app." So, that's how it's able

learning app." So, that's how it's able to retrieve that information, even though it's not in its core memory.md

file. So, I like I said, these two tiers, I guess like tiers, levels, whatever, levels of memory are actually really pretty good. But, if you really want to enhance the system, there are two things that you can do to push you

up two notches that can make it even better. So, that is tier number three.

better. So, that is tier number three.

You can install a external provider memory plugin. My favorite one to use is

memory plugin. My favorite one to use is called Honcho's. Honcho's doesn't

called Honcho's. Honcho's doesn't actually replace tier one and tier two of your memory. It just adds on to it.

It's looking at all your interactions with your Hermes agent and collecting what is called implicit understanding, patterns across your interactions and things you may not even be aware of about yourself, the projects that you do, the way that you behave, and it

supplements the Hermes agent with this information so it has better memory. And

to do this is also very easy. Cookie for

anybody that can guess. Yes, you just go into a new session. So, say, "Help me set up Honcho." And it will walk you through the exact steps of setting this up, including going on the Honcho terminal. I think you do get quite a lot

terminal. I think you do get quite a lot of free credits right now, at least at the time of this filming, so you can take advantage of this. Then, after you have it installed, it's just going to silently be working and collecting data, and you can have some conversations with

it. You want to see what's happening,

it. You want to see what's happening, you can look at the web app on Honcho, or you could just go to your terminal.

Okay. If you do know how to use terminal, this is very handy. Okay, so I have a three different terminals running at the same time, just like ignore what's going on. Let's focus just on this one. Ignore all of that. So, we can

this one. Ignore all of that. So, we can just write like Hermes Honcho status, and we can see, great, after it's installed, it's true, blah blah blah, it's it's done correctly, and then it's

telling you me information about myself.

Like, so here's my user Picard, it has information about this explicit observations, it also says some of the information that's observed. So, great,

we know that it's working. You will just find that as you're talking to your Hermes agent, sometimes it would start pulling information from Honcho when it needs to have specific information. It's

like a subtle thing, but you'll notice that it just makes your experience so much smoother. Would recommend this if

much smoother. Would recommend this if you intend to use Hermes a lot. Now,

finally, the fourth tier. You really

want to make your memory system super duper robust by setting up what is called a second brain. That is the Obsidian thing that I was showing you earlier. So, this is one of my Obsidian

earlier. So, this is one of my Obsidian second brains. And the way that you do

second brains. And the way that you do this, just go to Obsidian, and then click download, and you would be able to have it as a desktop app. And once you open it, you'll be greeted with something that looks like this. Click

open folder as vault, and you can just pick it wherever you want, just create a file, call it like Hermes vault, or whatever, and then just click open, and then you have your vault. That's

currently empty. What you need to do is you have to connect your Hermes to your vault, which, again, all you got to do is just ask your Hermes to help you set up an Obsidian vault. But basically,

after you do this, as I showed you earlier, when you're talking to your Hermes agent and you're doing research or whatever, you can ask it to use the Obsidian skill in order to save certain things into your Obsidian vault, and it can do this for you. Like I said, this

is just one of my vaults. I have like daily briefs that are here. I also like to save projects that are here. I'm

going to put on screen some other things I like to save into my vault, which could be useful for you. So, I like to save like certain projects. So, in this way, like say at some point I actually want to build like this AI learning app, right? Like all this information is

right? Like all this information is here. So, why is this useful? So, first

here. So, why is this useful? So, first

of all, you can like read the things in your vault, which is very nice. Uh maybe

you're like, "Wow, I wonder what it says over here." But on on top of that, your

over here." But on on top of that, your Hermes agent is also able to access this information that you stored into the vault. So, you can continue working on

vault. So, you can continue working on things from there. For example, I have like this food tracker PRD, uh which I made in which I wanted to track like the food that I'm eating because I'm too lazy to track my food, but I really

should cuz I'm trying to get more in shape, you know? So, I wrote this PRD for this like app that I wanted to build, but I didn't actually like ever build it. So, what I can say here is,

build it. So, what I can say here is, "Use Obsidian skill." Probably don't you actually need to like specify Obsidian skill, but I'm just going to do that just to make sure it's like very specific, "to grab the PRD for the food

tracker and come up with a step-by-step plan for how I can implement the MVP version in 1 hour." Great. So, you can see that it's using the skill view and

it's able to grab it. Okay, great. So,

it came back and now it has a plan. So,

it gives me the key MVP scope for 60 minutes. I should get like a photo of my

minutes. I should get like a photo of my food GPT-4 vision and Discord replies with macros, Obsidian JSON storage with the daily totals, and having specific commands like today log and goals I'm

able to see what it is that I have been eating and logging. And I should not have weekly chronic corrected flow because I want to do this in like 60 minutes, right? It tells me what the

minutes, right? It tells me what the tasks are and blah blah blah blah blah, etc. So, if I actually want to do this, then I can start implementing it.

Anyways, not going to do that now, but I hope you can see why having this fourth level of an Obsidian second brain is really, really helpful. Especially as

you do more work, you save more stuff, and there's more things that you would want your agent to be able to help you with. Okay, so there is one more topic

with. Okay, so there is one more topic left for your Hermes setup that will set the foundations so you can build so many amazing things. And that is open-source

amazing things. And that is open-source AI.

Now, I want to make a caveat here. This

is optional. If you are super baller and you're like totally cool just paying for like Anthropic Claude models or like other models through news for whatever and you're like willing to spend like hundreds of hundreds dollars, that's

totally cool, okay? That's totally cool.

I FELT THE STING AND RAN. But for me, I am not willing to spend that much money.

So that's why I started using local AI models. Models that I have downloaded

models. Models that I have downloaded into my Mac Studio, runs completely locally, it's private, and everything is like just is contained within this Mac Studio, which is actually kind of crazy, right? That you can just like be able to

right? That you can just like be able to own and use this like very powerful model. By the way, that is actually what

model. By the way, that is actually what I've been using this entire time throughout all of these like demos and stuff that we've been doing. This is a completely local model. Like when I use it, it's just on my machine. I don't

need to pay for anything. It's free. So

yeah, the one I'm using is a Qwen 3.6 35B model. And I'm actually just going

35B model. And I'm actually just going to put on screen some of the other local models that I do use for different things. Not going to go into too much

things. Not going to go into too much more detail about that. The general

consensus is that if you want to save money like running Hermes and agents, using local AI models is going to help reduce that cost substantially. And

there are some really, really good local AI models now that honestly like we were seeing me do stuff, right? And it's it's like just like really, really good. So

if you have the hardware and the ability of doing this, then I would recommend that you go for it. Now, the easiest way to use an open-source local model is to download Ollama, which is a model package manager. So you go to ollama.com

package manager. So you go to ollama.com and then click download. Then when you open it and you open the launch tab, there actually is Hermes agent here. So

you can just copy-paste this terminal command. Go to your terminal, paste it,

command. Go to your terminal, paste it, press enter, and you can select one of these to get started. It actually gives you recommendations. Or you can go to

you recommendations. Or you can go to the tried and true path now and just say, "Hey Hermes, help me use Ollama to download open-source models and use them

with you." Click enter and just follow

with you." Click enter and just follow the instructions. However, what I prefer

the instructions. However, what I prefer to do is use something called llama.cpp.

Without going into too much detail, llama.cpp is actually what Ollama uses.

llama.cpp is actually what Ollama uses.

It just like wraps around it, so it's easier for you to use. But if you use the underlying thing itself, which is llama.cpp, it gives you direct control

llama.cpp, it gives you direct control with the model, so you can tweak it to the way that you like it, and it's faster as well. Took me about like 30 minutes to set up, you know, again, using Hermes to help me like set it up.

Got to do like some terminal stuff, but you know, after 30 minutes, I was able to have my model up and running, and it's been great. Would recommend. All

right. Honestly, congratulations for making it until this part of the video.

I know we've covered a lot today, but trust me, if you have that patience to set it up like this, and just like do it properly, you're going to get so much more out of your Hermes agent. It is

amazing.

[music] So, for the last section, I'm actually going to show you what you can build once you lay this foundation. I'm going

to show you my multi-agent framework.

This is my Discord, where I primarily build workflows and softwares with Hermes. So, let's actually build

Hermes. So, let's actually build something together to show you how it works. Recently, I've been really into

works. Recently, I've been really into building these little desktop productivity learning apps. Like this

one, for example, it allows me to just input a study plan or whatever it is that I'm doing, and then I can like check them off, and you know, yay, it tells me that it's complete. And most

importantly, it saves all of this to my Obsidian vault. So, after I finish these

Obsidian vault. So, after I finish these plans, it would automatically save it, and I can actually see everything that I'm doing every day. So, let's build another one of these little companion apps today. I'm going to go here and

apps today. I'm going to go here and say, "@Hermes, draft the specs for a simple Pomodoro app that has the same style as the roadmap tracker app. I want

to be able to optionally input [music] what I'm working on, and then set the amount of time. I want the amount of time to be visual, so I can see

it decreasing, like a visual timer. I

also want things saved to Obsidian Vault as well. Click enter.

as well. Click enter.

Cool. So it looks like Hermes has picked it up.

Okay, great. It gave this PRD and it looks good to me. So let's say yes.

Queue it, please.

This says that it is queued and it will let me know when it's ready or needs an input. Sounds good. And hey, look at

input. Sounds good. And hey, look at that. Yay! Under the builds channel, we

that. Yay! Under the builds channel, we can see that the build is successful.

All right, moment of truth. Okay, so it has been built. Let's try it out. Let's

just say 1 minute and for the label, we'll just put test. [music] Click

enter. Okay. So just to see that it's actually floating on top. So if I go to like Hermes for example or some other apps. Okay, so we know that it's always

apps. Okay, so we know that it's always going to be on top, which is good. Uh

yep, and we can also move it around to the side as well.

Okay, that was so obnoxious. Let's go to Obsidian and there you go. 1 minute it's called label test and it's completed.

Yay!

So yeah, that is a full build uh using Discord. Since it is a multi-agent

Discord. Since it is a multi-agent system, you can also come back and ask it to do other things as well. Like hey

Hermes bot, draft the spec for some ideas of other desktop widget apps [music] that I can build.

Et cetera. At the same time, you can also ask it, Hermes bot, uh draft the spec to change the colors of the Pomodoro app to more pastel, for

example. Yep, and they're going to be

example. Yep, and they're going to be working in parallel. Yeah, isn't that just kind of like mind-blowing? Like

I've built so many different types of software and workflows using this multi-agent system. Once you get to this

multi-agent system. Once you get to this stage, Hermes also has a native orchestration via this Kanban board, which you you actually like orchestrate all of your different agents and coordinate them and make them like just super super optimized. So yeah, I've

been having so much fun building with Hermes. So if you guys want to see

Hermes. So if you guys want to see another video where I go more in depth about this multi-agent system and other projects that you can be building, let me know in the comments if you're interested in that. If you are, I'm very

happy to go more in depth into this. I'm

going to put on screen now some of the things that I have my multi-agent system set up, more advanced things and skills and just like things that I'm doing right now that you can do, you can unlock set up your Hermes agent properly. What I'm particularly like

properly. What I'm particularly like really into right now is actually using hardware and connecting hardware with my Hermes. So yes, super exciting. But for

Hermes. So yes, super exciting. But for

now, I think this is way enough to get you started. Thank you so much for

you started. Thank you so much for watching until the end of this video. I

hope this is helpful for you. I'm going

to put on screen now a little quiz that will help you retain all of this information that we've covered. And let

me know in the comments what you are planning to build with your Hermes. I

will see you guys in the next video or live stream.