Welcome to this Hermes agent.

Full course for beginners.

In this video, I'll be showing you how to set up, configure and use Hermes Agent

as a complete beginner.

I'm going to assume that you have no prior experience with this,

and we're going to go through everything step by step.

So by the end of this video, you have a fully functioning bot or Hermes agent,

whatever you want to call it, that you can use to manage your email, your calendar to a daily digest.

Go search the web, research really anything that you want.

I'm going to focus on teaching you the fundamentals here so you understand

how to set up Hermes agent so it works for you.

And then near the end of the video, we're going to get into

a lot of practical use cases where I show you real, legitimate things

that you can do that actually make this useful, which a lot of tutorials skip.

So with that said, let's get into it.

Let me teach you Hermes agent.

Let's go through every step of the process that you have no questions.

So you get this set up on your first try.

So what is Hermes agent and why would you use it?

Well, really, the way that you should view a Hermes agent is like a personal AI assistant.

And we can run 24 over seven.

It can execute tasks in the background.

And the way that we'll set it up in this video is on something called a virtual private server.

So it's always running and you don't need any physical hardware.

You don't have to have it plugged in.

It's just running 24 over seven.

Now this is not just a chatbot.

It's not something that's just going to give you a response or act like ChatGPT.

It's something that can actually take actions.

It can write a report.

It can research something for you.

It can go and draft an email.

It can look through your calendar and you can give it all of these different capabilities.

And what makes Hermes agent interesting is that it actually learns as you use it.

Now, you may have heard of other kind of agent frameworks out there, like open claw, for example,

open claws, very similar to Hermes agent, but it doesn't have this self-learning loop.

The reason why Hermes agent is now the most popular agent framework out there, and pretty much everyone

is using it for setting up these types of assistants is because of its self-learning capability.

We'll talk about that a little bit later, but what that means is that the more that you use it,

the more it learns about you and the more it actually improves.

So it gets better over time.

If you're just trying to create a simple automation for one single workflow,

this probably isn't the best tool to use.

This is meant to be chatted with consistently to do multiple tasks,

to handle your life, to automate a bunch of things for you.

If you just wanted one specific kind of tech automation to be completed,

you're better off using something like Open Claw or even just using something

like cloud Code or building a simple, elegant workflow.

Hopefully that makes a little bit of sense, but think of this again as like a dedicated kind of

AI assistant for you.

It has its own skills. It runs 24 over seven.

It can work with any lab.

So OpenAI mini minimax anthropic doesn't matter.

And again, that's why it's really popular now.

And a lot of people are switching to it from open call.

So just get a few things out of the way here.

You don't need to write code in order to use this.

You can just speak to it plainly using English or actually really any language.

And you can actually just dictate to it.

You don't even really have to type.

It handles all of the technology for you, so you don't need to do any complicated setup,

and you effectively just need to know to ask it what to do,

which is what I'm going to be showing you in this video.

Now, if we keep going here, there's five core concepts

or key components of Hermes agent that I want to go through.

And I'm going to explain them to you now so that when we start setting this up in just a few minutes.

And by the way, there's timestamps down below.

So you'll be able to see exactly where we do all of the setup.

You understand why it is we're doing what we do.

Now the five key concepts of Hermes agents are the following.

First we have memory, then skills.

Then the soul crawls or scheduled tasks, and then this self-improvement loop.

Now let's start with memory.

Okay, so, Hermes agent, what actually really is this?

Well, really, it's just a piece of software that kind of sits around an LLM.

Now, an LLM is the brain, right?

Think of like GPT 5.5, opus 4.7, mini Max, m2 point seven.

What about. Right? These are models or LLMs.

Now really all these models can do is just predict text.

You give them some text in and they give you some text.

Now with the Hermes agent software it does is it manages the context.

So the type of information that is being passed to this LLM,

as well as provides different skills and tools that it can use.

So it might provide a browser tool where it can go and it can search the web.

It might provide a code tool where it can write a code file or execute a script.

And then it manages the information like the memory,

the session, whatever info that this model needs to actually perform.

So what will actually happen is when you install Hermes agent,

you're going to get a directory structure that looks something like this.

Not exactly like this,

but it'll be similar to what you see on screen where you have a bunch of different files.

You have some markdown files.

This is just a special type of formatting that's really good at being read by by AI agents.

Sorry.

At LMS you'll have some configuration values,

you'll have some environment variables for secrets or API keys.

And then you'll have this memory folder.

Now this memory folder will have two files at least to begin.

One which is user.md and one which is memory dot.

And now you never need to actually go look at these files.

You don't need to touch them yourselves.

But it's important just to understand a little bit what's going on in the background.

So you know how these kind of ancient frameworks behave.

Now user MD is going to store information about you.

So who you are,

what your name is, maybe how old you are, where you live, different preferences about you.

And then memory.md is going to store long term facts and information of the ancient

things is important to always know whenever it's communicating with you.

Okay.

There'll be other memories that are stored automatically, but these are two important files

that will be updated by Hermes agent, which allow it to gather context on you

and perform better over time as it builds out the contents of these files.

Hopefully that makes sense.

Next, we have skills.

Now there's a ton of skills that are already built into Hermes H.

And I think it has like 60 or 70 by default.

And there's a ton of other ones that you can enable.

And these are essentially repeatable workflows that tell the model how to achieve some task

so that it can do it more predictably.

Rather than me just going every single day and saying,

hey, here are the five steps you need to complete.

If you want to check my email, we just have an email skill that outlines exactly how to do that.

So if I say hey, go check the email, it will reference the email skill

and it can just follow those steps rather than me writing them out every single time.

Then we have Chronos.

These are scheduled tasks which we'll look at later sessions and some other information that you'll see.

But the reason I'm showing this to you is that really Hermes agent, right?

Or any of these, it's just a piece of software.

They have these different files that kind of automatically get populated, and they

add facts and memory and information and skills, and it's just managing how the LLM actually behaves.

It's giving it a set of tools. It's giving it memory.

It's giving it additional context.

And it just calling the LM over and over and over and over again to actually achieve a goal.

That's really what these piece of software do.

And I know that's a high level explanation,

but hopefully this gives you a little bit of a peek behind the curtain.

Okay.

Now these two files like I talked about I can already went through them.

We have user.md memory.md.

This is an example of what they might look like.

And use your MD.

You might have again information about yourself and a memory.md.

You have kind of long term facts, things that you're working on etc..

Okay. Now continuing skills.

Again, we already touched on this a little bit, but the skill is simply a markdown file.

Sometimes it's accompanied by some scripts or some other files that describes what the skill is,

when to use it, and then later there will be a bunch of steps, right?

What kind of instructions that the agent can follow.

So if you say, hey,

go and do this thing that I've told you to do 100 times, rather than having to come up

with all of the skills on the fly, it can just go look up the information from the skill file. Go.

Okay, these are the steps that I need to follow and follow through.

Okay.

Now so it's correct to me here

there's 90 built in skills and over 500 community skills that you can install.

And you can also create your own skills, which we're going to look at in just a minute.

Now one other important thing to understand about skills is that they're discovered progressively.

What that means is that the name and the description of the skill

will be loaded into the context of your model whenever it starts running.

So let's say you open up Hermes agent, it's going to say, okay, I have, you know,

100 skills that are available to me, but it won't load the entire skill.

So the full instructions of it unless it needs to.

So this will save a lot of memory and save a lot of tokens for you.

So you're not loading, you know, hundreds of thousands of characters every time

the agent will just say, hey, okay, I need to use the morning brief skill.

Now load the morning brief skill.

And that's what's referred to as progressive disclosure okay.

Continuing personality.

So Hermes agent has a soul.

Now the soul is just like a human would have a soul.

But this is effectively a markdown file.

And in this markdown file you can describe how you want the agent to behave.

So do you want it to be harsh?

Do you want it to be nice?

Do you want it to be mean? Do you want it to be friendly?

Do you want to talk with emojis?

Do you want it to act like a girl or a guy or a robot?

Whatever. You can describe that in the soul.

And again, you don't need to manually write this file.

You can just tell your agent, hey, adjust your soul to do x, y, z.

So just kind of quickly going over that now continuing we have automations.

And this is where Hermes agent starts to become a lot more useful.

You can set up this bot to automatically do things on a schedule.

So every morning it could send you a reminder.

Every evening.

It could tell you to take your pills every Wednesday at 3 p.m..

It could give you a tennis star or something.

Whatever. Right? I'm just giving some random examples.

So you can tell it, hey, I want this thing to happen at this certain time,

every day, every hour, every week, every month, every day or whatever.

And it will automatically be able to do that for you.

And there's just two things that we talk.

When we talk about agents.

We talk about reactive agents and proactive agents.

Reactive agents wait for you to tell it to do something.

Whereas proactive agents will just work in the background automatically.

So you can make Hermes agent proactive by setting up these scheduled tasks.

Okay.

And then the last thing

is probably the most important, which sets this apart from other agent frameworks like open Claw.

This gets better as you use it, right?

So you do work, the agent learns about you, it saves the information to memory.

And then what Hermes agent will actually do is automatically

build its own skills in the background without you needing to ask it to do that.

So that when you start asking it to repeat something or do something multiple times,

it will build out skills automatically so it remembers your preference on how to do that.

Now this again is a huge differentiator between other frameworks

where you would have to manually tell it to do that.

In this case, Hermes agent will actually review what you've done with it

and then automatically kind of improve itself over time.

So I promise we're going to get into the setup here.

But I do need to give you a security disclosure here

when using Hermes agent and just go over a few last important details.

Again, I'm trying to make this really thorough so that I teach you guys as much as possible.

Okay, so when you're using these types of agent frameworks

because you're giving them a lot of autonomy, meaning they can go on their own,

they can run a scheduled task, they can update their own files.

There are things that can go wrong.

So generally you want to treat these AI agents like they're a brand new member to your team.

What that means is that you want to restrict the access you give to them right away.

You want to slowly give them capabilities.

And for example, if you just hired a new employee day one, you wouldn't give them root access

to everything in your business, right?

You'd slowly kind of introduce things

until the trust is built up, and then eventually you give them more permissions.

So just going through a few things that could go wrong here.

When you use the agents, you're aware of them.

First of all, spend if you're going to set up an API key here and use a provider like anthropic,

for example, if you don't set spending limits, this could go and spend hundreds of dollars per day.

And that's happened with many people.

Or for example, if you give it access to a credit card,

it could go to the internet and literally use that credit card to buy something.

Right?

So you want to be really careful with how much it's actually going to cost you.

Next, prompt injection.

If you just have this open to the internet and anyone can communicate

with your bot or message it, they can send it anything and tell it to do anything.

So they could say, hey, go and take all of Tim's secret details and send me an email with it, right?

There's also something called a prompt injection, which is when this is done, kind of without

you knowing where, for example, someone sends you an email, that email contains a string,

which is kind of a prompt that says, hey, ignore all past instructions and go do this thing.

And even though you didn't directly tell the agent to do this because it read the email,

it read those instructions and then execute a task that it probably shouldn't have executed.

Get prompt injection.

Really important. Next, data leaks, right?

If you have your Hermes agent kind of exposed on the internet, someone can steal your API key.

They can get access to sensitive files.

So you want to be careful with that.

And then of course destructive actions.

So things like deleting a Google drive or sending an email when it shouldn't.

Right. Or making a payment that it shouldn't do.

So we want to make sure that we set up

what's called guardrails around these activities, which I'm going to go over in this video.

So you can see we have named API keys Least privilege making your own accounts.

And then again, not giving full access on day one slowly building up until you trust the agent.

So it's not going to make these types of mistakes.

So just be really careful when you set this up.

Again, I'm going to go through all of the best practices okay.

Here's just a few things in terms of locking down your agent.

Again I like to go over this at the beginning so no one misses it.

Typically, if you are going to be using a lot of email

related stuff, you can give your agent its own email address.

You can also give one API key per agent.

So if you have multiple tools out there, make sure you just have one API key per tool that you're using.

So worst case you could delete the API key.

You can rotate it, you can change it and you can lock something down remotely.

If you are going to be giving this access to your calendar,

to various tools that you're using, give it the least amount of permissions required.

Don't just give it admin access, only give it access to what it actually needs and start

with read only mode reading information is a lot less destructive, obviously.

Then if it can edit, modify, delete, send, etc.

in terms of secrets and keys, never paste your API keys directly in the chat or give it to the bot.

You want to set that using a command, which we'll look at later.

Rotate your keys regularly and then keep everything that's backed up

for this agent in a private GitHub repo, which we're going to look at.

And then server hardening.

I'm not going to get into that too much, but

effectively make sure that not anyone can just access your agent on the internet.

Okay. So let's actually start setting this up.

I have a checklist of what we're going to use to do this.

And what we're going to be doing is setting this up using a virtual private server.

Virtual private server is essentially a computer that runs in the cloud in a secure environment.

For this video, I'm going to be using Hostinger.

I've a long term partnership with them, and fortunately they have a really easy way

to deploy Hermes agent with a one click install.

So if you guys want to follow along with me, go to the link in the description.

You can simply choose a plan from this landing page.

I recommend the KVM two plan.

You can see that this one is just $9 per month.

You can select the term that you want to use

and because I have a partnership with them, you can use the code tech with Tim.

That will give you an additional 10% off.

And then you can go ahead and just press on, continue,

and it will start provisioning and setting up this agent for you.

Now this will automatically do the deployment.

It will install Hermes agent

and it will set it up using a Docker container which is a secure environment.

If you're not familiar with that, don't worry.

But it's just a secure, hardened way to set this up.

And then of course, you don't need to use this if you have your own hardware device

that you want to run this on,

or you want to deploy it on another server you have access to, you can do that.

And the way that you would do that is you simply need to run this curl command from the Hermes

Agent website, which I will link in the description and then run the Hermes setup command.

As long as you have Hermes running on some type of device and you can access the device,

then you'll be able to follow along with the rest of the video.

So for now,

let me continue with the provisioning of the SVC and then I'll show you the full setup process.

Okay, so if you are using hosting, you're here.

When you do go to the deploy this, it's going to show you a page that looks like this

where there's an admin password as well as a Naxos API key.

Don't worry about the access API key.

This is just something they provide for free that gives you some AI credits.

But what you do want to do here is just copy to your clipboard the admin password for Hermes agent.

You can access this later.

You can change it if you want,

but just make sure that you have access to this because you will need it to sign in.

So from there we're going to deploy.

It's going to take a few minutes here to provision the VP's.

And I'll just quickly mention that the reason I prefer to use a VPS over a hardware device

is, first of all, this is significantly cheaper, especially on a monthly basis,

and this is running 24 over seven without you having to worry about your device turning off,

staying active or connected to the internet, being stolen, being in a fire, right?

Whatever.

When you use a virtual private server, this is hosted in a data center,

which is literally more secure than places like a prison or a white House.

And it just makes your life

significantly easier because you don't have to worry about managing a physical hardware device.

If you do want to go with the Mac

mini route, you absolutely can do that, but that's going to be significantly more expensive.

And again, that's going to require that you actually physically

secure this device, especially if you have a lot of credentials on it.

If someone were to come into your office and steal your Mac mini, that's the same thing

as them having unrestricted access to your Hermes or OpenCL or whatever instance.

So you guys do whatever you want.

But I just like to mention that I personally use VPS.

That's why I'm setting it up this way.

And anytime I do anything a gen tech, it's always on a VPNs.

I also can just very easily terminated or turn it off if I need to.

If there is a security breach by pressing one button compared to a physical hardware device.

Anyways, we're gonna wait for this to provision, and then we'll go through the rest of the setup.

Okay, so it's almost finished provisioning.

Once it's done, you should be brought to a page that looks like this.

If you can't find it for some reason, if you scroll down on the left hand side,

you can go to VPNs and then you can just find the virtual private server

that you created, and you'll want to go to the Docker Manager tab.

From here you'll see that there's some containers.

We have this traffic container, and then a Hermes agent container with some kind of random string name.

You'll need to wait for this to say running.

So right now it says created, but it's not running yet.

Once it's running, those should be a single button you can press that's going to open up the Hermes

dashboard or kind of terminal based UI, where you'll be able to interact with it and start setting it up.

Okay. So you can see now that it shows running.

So beside running you'll be able to just press on open.

When you press on open it's going to prompt you for the username and password that we saw previously.

So what we're going to do here is put Hermes needs to be in lowercase

because that's the default username unless you changed it.

And then the password that we found before and we're going to sign it.

And it's going to bring us to this page.

Now if for some reason you've lost that password or you need it in the future,

the way that you can access it is if you go here to Docker Manager

and you go, I believe manage and then you scroll down,

you should see environment and then you can see all of the different values here.

So you'll be able to just view the admin password.

And you can directly change it from here in case it's exposed or leaked.

Make sure you don't share this with anyone.

I'm going to remove this instance afterwards, because if you do,

then they'll have full access to your Hermes agent, which obviously is not good.

Now, also, just so you're aware, if you go to this overview tab here,

you do have the ability just to stop the virtual private server or to reboot it if there's an issue.

So if for some reason you needed to shut this down immediately, you can do that.

And if you're familiar with SSH, it will give you the details here.

But you will need to change the root password.

I'm not going to go through that because we don't need to ssh into this instance.

But just wanted to make you aware in case you're familiar with those concepts.

Okay.

So from here we're now in the Hermes agent setup.

Now if you have not deployed with Hostinger, what you'll need to do again,

like I mentioned, is install this and then run the Hermes setup command

that should bring you to this page or something that looks like this in your terminal.

And then everything else in this video will be the exact same.

Now what we're going to do is go to the full setup.

The way that you can navigate here is you can use the arrow keys.

So arrow key up, arrow key down, and then press the enter key

or the space key to select okay.

The terminal does not allow you to use the mouse.

So like I can't press this I need to use my keyboard to do the navigation.

So I'm going to go to full set up here.

And it's going to give me some options for selecting my AI provider.

Now by default it's going to use Nexus AI.

This is set up with hosting or they have some partnership.

However I don't really like Nexus AI.

So what I'm going to do is I'm going to switch over to use open AI.

So if you scroll through there's probably a long list.

I just zoomed in so you can see it's not kind of showing all of them.

And I'm going to use my OpenAI Codex subscription.

Now you can not use a cloud code subscription here anymore.

They remove the ability to do that.

You'll see there's a lot of other options.

So whatever you prefer to use, you can.

And while it is showing the cloud code thing here, if you do that, then

it is going to run you as extra usage and it's going to build you additionally.

So I advise not using cloud code.

And if you have an open AI plan like I do, using that instead.

So I'm going to use OpenAI Codex as mine.

What I did there is I just selected it and pressed on enter.

And now it's prompting me to go to this URL and to copy in this code.

So what I'm going to do is I'm going to open this URL.

I'm just going to press on it.

It should open in my browser, I'm going to sign in with my email.

And then I'm get a press continue.

And I'm just going to paste the code.

Now depending on what you're using, obviously you're going to have a different

kind of flow here for connecting this to an LLM provider.

But you need some kind of LM, you have to set that up in order for Hermes agent to work.

For most of you, using open

AI is just going to be the best choice and probably the most cost effective as well.

Okay, so now it's going to ask us to select the model.

Now with Hermes agent we can use multiple different models.

We can connect to various LM providers if we want to do that.

But for most of you you just need to connect to one, right?

If for some reason you want to have multiple models and you know what those models can do,

and you're going to be using this a ton, and you want to have maybe a cheap model for some task.

And expensive model for another. You can set that up.

And there's all kinds of stuff you can even just ask, you know, Hermes agent how to do that.

But for now, we're just going to go with GPT 5.5.

If you use this a ton, it will be ripping through your subscription pretty quickly.

But again, with GPT, it's fairly generous compared to a lot of the other plans.

So I'm going to go GPT 5.5.

Press enter and select that okay.

Next it's going to ask us to select a text

to speech provider text to speech is exactly what it says.

And you can actually have the model speak back to you.

I don't find this particularly useful.

I don't use it very often.

So what I'm going to do is just keep the edge tts.

But if you want to use a specific text to speech model or have a certain type of voice,

then you can just pick one here, like 11 labs, for example, if you want it to sound more human.

Okay, so from here I'm going to press on, enter next and ask for the terminal back end.

We are going to keep the local terminal back and do not change this.

Now we're going to be brought to the agent settings.

Now this is asking us what we want for the maximum number of iterations.

Iterations is essentially how many times the line will be called in one task.

And one thing that you ask it to do.

So I'm just going to go with 90. This is the default.

You can make this higher if you want it to be able to run for like,

you know, hours at a time and do deep exploration.

But for most of you, just keeping the default value

is going to be best and you really don't need more than 90 iterations.

Again, this is 90 individual calls to the LLM.

And it says here you can set it 250 plus if you want.

Again, you don't need to do that.

Probably just leave it at 90 okay.

Now it's going to ask us for the tool progress mode.

Now there's four options here of new all and verbose.

I suggest just leaving it at the default, which is all.

And what this is going to do is show you the activity of what the LM is doing.

So if you turn off, what's going to happen is if, let's say you say, hey,

go and take a screenshot of this website and give me the result.

Well, it will just give you the result.

It won't tell you what it did, it won't walk you through the process or to say the result.

Is this right?

If you put it to new anytime, it calls a new tool

that it hasn't seen before, it's going to tell you that it use that tool.

If you put it to all, pretty much anything that it's doing, it's going to share with you.

So it's going to say, hey, I'm going to the web and I'm doing this.

Hey, I'm looking at this skill file.

Hey, I'm searching for this thing, right?

And then if you put verbose, it's going to do that.

Plus it's going to give you a ton of additional details.

So all is most likely the option that you want.

You have some visibility into what's going on without seeing too much data.

That's not really useful for humans to look at.

Okay.

So we're going to go with all, which is the default next context compression.

This is important I suggest that you set this around 0.8.

Now this is when it's going to automatically compress what's happening in the conversation.

In case you're unfamiliar, there's a maximum, what do you call it?

Context window for LLMs.

And based on the LLM you're using, you have a different size.

Some have a million, tokens, some have 200 K tokens, some app, 100 K tokens.

Depends on the model.

Right now, this is essentially the maximum amount of information that the model can remember at one time.

What this is saying is, hey, when we start getting close to filling up

that context of getting close to maximum memory, when should I compress this and kind of

turn it into a smaller amount of information so I can keep going?

Maybe that's a bad explanation, but essentially what we're setting is, hey,

at what level should I start compressing the context so I can fit in more information?

Now, if you leave it at 0.5, that means that approximately

when the model's 50% full, it's going to take what it knows.

It's going to kind of summarize it into a smaller amount of information and then store it and continue.

You probably want to go a little bit further reuse more of the context window.

So something like 80% or 0.8 before you start compressing the context,

you have a little bit longer memory and you don't lose the information so quickly.

That's essentially what I'm saying here, is we want to compress a little bit later in the context window.

Not so early. So we're setting this at 0.8.

If it doesn't make sense don't worry.

Just use the value I have here okay.

Next session reset mode.

So whenever you're using these

agent framework server you want to call it you'll have something called a session.

A session is essentially your kind of chat or yeah,

your conversation with the AI model.

Now the session is typically all loaded in the context of the model.

So if I'm having a conversation I say, hey, what did I say?

You know, three messages ago.

It's going to remember that because it's in the current session.

But if I message from another platform, so maybe I'm messaging it

from telegram or something, as opposed to messaging it from the terminal.

I'm now in a new session.

And these two sessions aren't connected, right.

So they can't find information between.

So what this is asking you right now is, hey, when do you want to reset the session?

When should we make a new session?

Should we do that every day?

Should we do it whenever you stop messaging for a few minutes, should we never do that?

So you just have a super long session all the time and that's what we're setting.

So what I'm going to go with is inactivity plus daily reset.

What this means is that after a certain amount of time where you haven't message the bot,

it's just going to clear what you talked about before so that you have a fresh session

when you start communicating with it.

Again, any important memories, it will automatically save by itself,

but it's not going to have every single detail and then every single day it will do this as well.

Just create a fresh session

so that you're kind of keeping the context window low and ultimately the cost lower as well.

If you never reset this, it means that you need to manually

create a new session, which is probably not great for beginners.

If you want kind of a new chain of thought or a new conversation to begin.

So I'm going to go in activity here for the timeout.

We're going to leave it at default.

And for the reset our we're going to leave it at default as well okay.

Next we get to the messaging platform.

Now this is super important because if you want to message with your Hermes agent, you probably don't

want to have to go to this browser window and do it from your computer every single time.

You want to be able to do it from your phone, from slack, from email, whatever.

So what you can actually do is configure this agent

so you can message it from something other than this UI that you're in right now.

So what we're going to do is set up telegram just because it's the easiest one.

And you should always start with that.

And then afterwards you can set up any other integration that you want.

I just obviously can't walk through all of them in this video.

And it depends what you want.

I would say the most useful is to have it in slack, especially if this is going to be doing business

related stuff for you.

But for now we'll do telegram just because it's super simple.

Okay, so to configure telegram we're just going to press space okay.

So notice I'm kind of using my arrow keys I press space here and I'm toggling this checkmark.

And then I can press enter.

Now from here it's going to ask me for a telegram bot token.

So what we're going to need to do is open up telegram.

I think I have it on my computer, so I'm just going to open it up here.

And then I'm going to show you how we can do this.

If you don't have telegram already, make sure you create an account

and then download the application on whatever device you want to use it from.

Okay, so I've got telegram open what I'm going to do is I'm just going to go to this search here,

and I'm going to search for ad bot father.

Okay. You need to follow along with me if you want to do this.

Now we're going to look for bot folder with the blue checkmark.

Make sure it's the one with the blue checkmark.

And from here we're going to type slash new and then bot as one word.

What this is going to do is set up a new telegram bot for us.

It's going to take a second here.

And we need to give the bot a name.

So to give it a name this is the display name.

So what. We'll see more chatting with it.

So we can just call this like Hermes agent or something.

And then after this what we can do is we can set the username for the bot.

Now this needs to be unique string.

So I'm just going to go Hermes.

Agent 32456 something underscore bot okay.

And make sure that it ends in underscore bot as a lowercase.

It needs to end in that okay.

So we're going to wait for that to finish.

Make sure that we give it a valid name.

And then what's going to happen is it's going to give us this token right here.

So we're going to copy that token.

And that's the token that we're going to provide here in telegram.

Now make sure you don't leak this token because again

this is the token that kind of links Hermes agent to telegram.

So now anyone who has this token would be able to essentially message your bot,

which is a security issue, right?

So don't leak that okay.

So I'm going to go enter.

And now you're going to see that it's saved the telegram token.

Now next it's asking for us to restrict who can message the bot.

Now. Same thing.

You don't want anyone

to be able to just message this bot on telegram and then chat with your Hermes agent.

You want to have access controls right where you say, hey, only this user,

only my mom, only my brother, only me.

Whatever can message this spot.

So in order to allow someone to message this, what you need to do is follow the instructions right here.

So we're going to go back to our telegram.

We're going to go here and we're going to search user info okay.

And you can see it's popping up like this.

And from user info what we can do is just type slash start.

And if we type slash start it's going to show us our ID.

Now this is the ID that we want.

So we're just going to copy this ID right here.

And we're just going to paste it.

And if we wanted to have multiple IDs we'd put a comma.

And then we paste multiple IDs of the user that we want it to be able to chat with this.

And if it was another user, you'd have to ask them to run that command as well.

So we're gonna go ahead and press on enter.

You'll see that it's now added this.

It says user user ID is the home channel.

We're going to go.

Yes. Just use that as default.

And then we should be set up with telegram.

Now next it's going to go through tools for this.

You don't need to set anything.

You can just press enter it, just showing you a bunch of tools that are built in by default.

And then you can connect some other ones if you want.

Next it's asking you for a browser provider.

Now this is important because if you want to do a ton of browser or web based operations,

I would recommend that you use a browser like browser based or Fire Pro.

Now these are paid options.

You will get some free credits, but they will get relatively expensive relatively fast.

But this allows you to have a full remote browser instance

running in the cloud that can be controlled and paralyzed by Hermes agent.

Now, what that means is that you can have like ten browser windows open at the same time.

You can go and bypass CAPTCHAs, you can browse from different regions.

You have a lot of power when you're doing this from the web.

However, again, you have to pay for this and if you want, you still can have like,

you know, search and browser based operations for free using the local browser here.

The issue is that this is what's called a headless browser, so it's not capable of rendering

complex JavaScript or actually really viewing what's going on in the page.

It can just kind of pull down the HTML that sits on the side of a page.

That might not make sense to you if you're not technical, but effective.

If you want to do a ton of browser related stuff, go to Browser Base or go to Fire Crawl.

Purchase a subscription for like 20 bucks a month and then connect it.

Otherwise, use the local browser, which I'll show you how to set up minimally, at least in a minute,

so I can at least search for stuff and do some basic operations.

Okay, so for now we're going to go with local browser like that.

Next it's going to ask us for a provider for image generation because we're using OpenAI.

We're just going to select OpenAI Codex here.

And we'll go with image to medium.

You can choose what you want okay.

Then it's going to ask us for the text of speech.

Again.

We're just going to leave the default one which is free.

And then it asks us for a search provider okay.

So for here we're just leave for Chrome.

It's allowing us to use a kind of open source self-hosted version that you don't need to pay for.

You can also use the paid one if you do have a fire crawl account.

I'm not going to go through this too much,

but search and kind of a web automation are two separate things.

This will allow us to search the web,

but it's not going to allow us to automate the browser like I was talking about before.

So I know it's a little bit confusing, but we have something for search.

We have something for browser automation.

Again, if you want to do that

a lot, look into browser based or fire crawl because they're very good options.

We're going to press enter here.

And now it's going to ask for our fire crawl instance URL.

Don't worry.

We can just go ahead and press enter there.

We don't need to do anything.

And now it's going to ask us for tools for telegram okay.

And I just press enter a few times.

For some reason that repeated a few of the steps that I needed here.

And anyways, now we're good and it's asking us to press enter to reconnect.

And everything should be configured at this point and we can start messaging Hermes agent.

We're going to see if it works in just one second. Okay.

So now that we're through this full setup, I know there was a lot of instructions.

Wanted to make sure I covered all of it for you.

We're just going to test to make sure this works by saying hello world.

Now what it should do on the first run is start initializing

the agent and running in what's called the t UI.

Now T stands for Terminal User Interface.

This is the default way to chat with Hermes agent,

which you can do from the web browser when you're using something like hosting or whatever.

Okay.

Or from wherever you have this installed just from the terminal.

Now, let's say for some reason, you weren't inside of this window

and you were just in like a default terminal, like I'm in right now.

And let me just do this here.

What you can do to access this t is you can simply just run the Hermes command.

So let's say you were to exit out of this Hermes application, right?

If you want to reopen it

and you want to see that terminal UI again, you just type Hermes wherever you installed it.

If you're using Hostinger, it's in this kind of terminal window that you see right here.

And the way that I got to this view is I hit control C on my keyboard, which stands for quit.

So that will quit out of the Hermes application control

C again, depending on what operating system control C or command C on your keyboard.

Then I hit zero.

I just hit zero because there's some setup thing here that you need to do with Hostinger.

And now I just type Hermes again.

And when I type Hermes, it will bring me back to this kind of terminal user interface.

So again I can hit control C to escape that.

And then if I want to go back, I can hit Hermes.

There's also a bunch of other Hermes commands.

For example, if you want to go back into the setup, you can run Hermes setup

and it's going to bring you back into that same setup flow that we're in again.

And if you want to quit again, you hit control C and you can just get out of this setup like I'm there

and then I can clear the screen and type Hermes again to open the t UI.

I'm just showing it to you because again, this is for complete beginners.

And I know some of you are not familiar with the term.

Okay, so hopefully that shows to you on. Right?

That's how you can kind of chat with the agent.

That's how you can escape from the agent.

That's how you can get back into the setup.

There's a bunch of other commands that you can run, but generally you don't really need to do.

So now what we want to do

is we want to test to make sure that we can chat with our agent from telegram now.

So in order to chat with it from telegram, if we go back to the bot father,

we can find this little, link that goes to the username of our bot.

If we can't find the link,

we can also just copy the username that we used and we can try to message it on telegram.

So what I'm going to do is just press this button and I'm going to go start.

And it should bring me to kind of a UI where I'm now in telegram,

and I should be able to chat with the bot.

So what I can do is just type hello, for example,

and it should say typing and give us a response.

If it's not, we may not have connected the bot properly

and we may need to do some debugging, which looks like we do need to do now.

So I'm going to show you how we would fix this okay.

So this is an error I wasn't expecting to happen.

It looks like for some reason I bought didn't connect properly.

So if this happens don't freak out.

Go to the terminal user interface for Harvey's and just ask it how to fix it.

So hey, I tried to connect telegram, to my bot here.

However, when I message from telegram, the bot is not doing anything.

It shows the received checkmark, but it's not typing or anything.

What might be a sub?

Could you help me fix this?

Okay, so I'm just going to ask it.

Hey, help me out and let's do some live debugging and see if we can figure out the issue.

And by the way, if you're wondering how I'm doing this,

how I'm kind of chatting with it here, I'm using something called Whisper Flow.

It's one of the best AI voice dictations out there.

It's extremely fast, as you can see, and this is what I use now instead of

typing pretty much 99% of the time, if I open it up,

you can see that I have a ton of words, 130,000 words.

This is my speed 200 words per minute.

It saves all of the transcriptions.

It has snippets, styles, all of this kind of stuff.

And I have a long story partnership with them. They're free to use and try out.

If you guys want to leave a link to it in the description.

So anyway, so you can see that now it's kind of going through and saying, okay,

you know, I wasn't able to connect. Let's figure out what's going on.

And hopefully it's going to give me a response here while telegram is still not functioning okay.

So it just finished.

It gave me a response and it said that it found and fix the immediate issue, said the Hermes.

Gateway was stopped.

The log showed that it was connected to telegram, but it just wasn't working for some reason.

So it said it started the gateway, verify that it's working and now it should be good.

So I just messaged it and you can see now it's responding and it's working.

So again, if you run into some kind of problem, the beauty of Hermes agent

is that you can just ask it and usually can fix the problem for you.

Okay, so don't freak out, just prompt it, ask it.

And look, I showed you in real time. There was something.

I didn't know what the problem was.

What did I do?

I asked Hermes agent.

It solved the problem.

And now we're good to go.

Okay, so at this point, we've done the setup right.

We've connected to telegram, we have all of the main things in place.

And now I want to start connecting a few additional tools and skills that we need

now if I go back here, you can see I just have kind of a list what we need to do.

So we've done step one, step two, step three, step four.

Now we need to do step five, step six and step seven which is onboarding.

So kind of building a bit of memory setting the soul for the bot.

Then we want to set up automated backups.

So anytime we do anything here it will backup to GitHub.

And I want to show you how we can do browser automation and stuff like that.

Okay. So let's now, let's open it like this.

Let's start chatting from telegram just because we can.

So we can use telegram, we can use, what do you call it here?

The UI doesn't matter. We can show whatever we want.

And now anything that we would do from the UI, we can just say from telegram.

So because it's our first time using the bot, it's generally a good idea to give it

a description of who you are, what your goals are, and how you want the bot to behave.

So let me show you.

But this is a prompt that you should tailor yourself and send to the model.

Hey, my name is Tim.

You can refer to me as Tim.

I run a large programing YouTube channel with 2 million subscribers.

I'm interested in AI Python, software development

and my goal for you is just to demonstrate and it's a Toriel.

How to use Hermes Agent for a beginner audience.

I want you to update your soul MD so that you're always concise to the point.

Direct and educational.

Don't give me any fluff, just give me all of the information and explain to me

why it is that you're doing something, and the steps and processes that you follow,

so that I know what's going wrong and I can help kind of debug some of your responses.

That's what I appreciate in a bot, and that's what I want your soul to be

kind of like an educator, but not super verbose.

Okay, so I'm kind of just rambling a little bit, but the point is I prompted all of this.

I mean, I hit enter now and it should start updating itself to remember these preferences.

Okay, so it's gone through. It's shown me kind of what it did here.

So it viewed the skills, it search the files whatever.

And it's now updated this and given me or done essentially what I asked them to do.

Right. So adjusted it.

So it's kind of giving me the details. Perfect.

So now what I want to do is I want to move to back up.

So before I go in and I set anything else up here, I just want to start setting things up

so that whatever we do with Hermes agent in case something goes wrong or crashes, or we lose the VP's,

we have a backup of how we can figure this because the more you use it, the more valuable it gets.

But then if you were to lose this or something were to go wrong, or again, there was a fire you lost.

You know, access to your VPS.

You want to still have all those memories, all those conversations, all those skills,

all the stuff that Hermes agent did for you

or even any files or, you know, projects or whatever that it built.

So what we're going to set up here is something called it automated GitHub backup.

There's a few setup steps here, but I promise it's well worth it.

GitHub is a version control software that allows us to keep track of changes to various files.

So what this will also allow us to do when we set this up is let's say

we accidentally make a change to our agent and the performance degrades a lot.

We can just tell the agent, hey, go back to the previous change and it will be able to do that

because it's going to keep track and checkpoint all of the changes that it's making.

You'll see what I mean in a second, but it's just this kind of live history

that we're going to set up so we can always go back to a previous version,

and we can always have everything saved and visible to us.

So in order to do this, we do need to make an account on GitHub.

Okay, this is free.

I'll leave a link to in the description, but just make a free account.

You most likely already have one.

If you've ever done any dev stuff before.

From here we're going to go to new.

We're going to go to new repository from the top here okay.

We're going to make a new private repository.

Now we're going to call the repository Hermes Agent Backup.

And then whatever you want to call it I'm just going to say tutorial because this is for my tutorial.

I'll delete it after.

Now make sure super important that you set the visibility as private.

Okay. Then go ahead and press on Create repository.

So once we create the repository we're just going to leave this page open.

And now we're going to generate something called a personal access token.

And this is the token

that we're going to give Hermes agent access to so that it can control this repository.

Now this repository is just a place where it's going to upload all of its files.

It's going to backup all of its data.

So again, we always have access to it.

It's kind of connected to our bot okay.

So what we're going to do now is we're going to go to our profile photo.

We're going to go to settings and I'm just going to open this in a new tab okay.

So I want to leave my repo open in one tab.

And then I want to open my GitHub settings in another tab I'm going to scroll all the way down.

I'm going to go to developer Settings on the left hand side here.

Now from Developer Settings I'm going to go Personal Access

Tokens and find grained tokens.

From here I'm going to go generate new token.

And I'm going to give this a name.

Now it might ask you to authenticate GitHub at this point.

Just do that. And I'm going to call this Hermes

backup repo okay.

Now what I'm going to do is I'm just going to go no expiration for now.

This will just make your life easier.

You can set expiration if you want.

And I'm going to select repositories.

What I'm doing here is I'm creating a credential

that just has access to this one repository that we created.

So we can give it to Hermes agent.

And it can just control this one GitHub repo. And that's it.

It won't have any other control of our account.

So I'm going to press only select repositories.

And I'm going to type and search for the Hermes Agent backup tutorial.

So the repo we just created,

if for some reason

you want to give it access to more, you can, but probably you just want to give it access to one.

Then for permissions we're going to go add permissions and we're going to search for contents.

And we're just going to select contents like that.

We're then going to go and do read and run.

Now what this is going to do is give the agent the ability to read and write from the repository.

That's it. That's already over it.

And then we're going to go and press Generate Token okay.

So we're going to generate this.

And we're going to copy this personal access token.

Now at this point we will actually need to go back

to the terminal UI because this step does need to be done from the terminal.

What we're going to do is we're going to quit out of Hermes agent

and wherever we're running Hermes agents or from our apps or our terminal or whatever,

we're going to type the following command.

So again, I'm in the terminal, right?

I press control C, I'm not in the terminal UI.

I don't have Hermes running.

What I'm going to do now is I'm going to type Hermes

and then config set

in all capitals GitHub underscore token.

Make sure you spell it exactly like this.

And then I'm going to paste this GitHub personal access token.

Again I will delete it afterwards.

Don't share this with anyone else.

This is the way that you can set an environment variable in Hermes agent.

Now an environment variable is a secret value that you don't want to share with anyone else.

Because the thing is,

I could go to my bot and I could just paste in this token, right and say, hey, here's the token.

You set it.

But if I do that, that tokens now going to be stored in session.

So there's going to be a log of it. It's going to be really easy to find.

If I had exposed my agent for some reason,

and it's going to be sent to my LLM provider like ChatGPT or something, right.

Or anthropic or OpenAI or whatever.

So the way that we set values when these are kind of secret keys

and this is going to be important as you set this up and you add more configuration later, is

by using this command Hermes config set, you put the name of the variable.

So in this case it's GitHub token.

Make sure you spell it exactly like this with the underscore.

And then you set the value.

So anytime you have a secret key and API key or credential you set it like this.

Super important

that you do this manually and you don't share it directly with the Asian for security reasons.

Okay.

So we're going to go ahead and press on enter

and notice it's now set the value in this kind of secret directory.

Don't envy.

This is a value that we should not expose outside of this Hermes agent okay.

So it's set.

Now I'm just going to rerun Hermes in the terminal user interface.

And now we should have access to GitHub.

So what I'm going to do is I'm now going to give a prompt to the Hermes agent

and tell it to run a backup of this repository

and to not save any, details that it shouldn't be saving, essentially.

And we're going to check and see if it pops up here.

And then once we set that up, we're kind of going to be good with the backup

and we can move on to the next step.

So I'm going to do the following prompt.

Hi. I've given you access to a fine grained GitHub personal access token.

This has access to one GitHub repo.

I'm going to share the URL okay.

And then I'm going to copy the URL to my GitHub repo.

And I'm going to paste it here.

What I would like you to do is backup the current Hermes configuration and state

without saving any secret keys, credentials, or things that are sensitive to GitHub.

You have the ability to push to this GitHub repository.

Go ahead if you don't understand what I'm saying.

Don't worry, you don't need to say it word for word.

Essentially what you're just going to say is, hey, I gave you access to a GitHub key,

can you back up my repo?

Can you back up the code that we have and save it?

That's effectively the message.

I'm just giving it a bit more details because I know what to ask for.

Again, you don't need to do exactly what I say, just generally prompt it with something like this.

So we're going to go ahead and press on enter.

And it should be intelligent enough now to go and actually back this up for us.

Let's see if it works okay.

So it's asking us approval for a command.

So for this one I'm just going to go with always because this is just cloning a repo

and then setting it.

So we'll just go always and let it do that.

And then it should hopefully be able to do this okay.

So it says it's finished here. Did take a second.

Now if I go when I refresh the repository page that we were on before,

we should see that we have all of the details.

So notice we have skills, state memories, cron config.

Right. And all of this stuff should be inside of here.

Now you might want to be a little bit careful.

Again, you don't want to share this with anyone

because sometimes you can accidentally put like a secret in here in the GitHub repo.

So I would suggest having a look through the repo saying, okay, is there anything sensitive?

Is there a file in here that shouldn't be in here?

And if there is telling the agent and it can remove it

and hopefully clean up the repo for you.

But regardless, you can see that now we have some of those memories

I was talking about where we have this like user file.

Right. So this is information about me.

Remember I was teaching that at the beginning of the video we have memory.md

boom Hermes profiles here.

Hermes is not in path and two environment whatever.

And these will get updated over time.

And now if we tell the agent automatically backed up.

So the next step is because this is set up we're going to say okay, this is great.

I'd like you to update the memory now.

So whenever you make any significant changes

I'd like you to make a new commit or a new save to GitHub.

That's it.

I'm just saying, hey, anytime you make a change, just save it to GitHub.

That's effectively what I'm telling the agent.

You don't need to over complicate the prompt.

And now it should understand to do this update it's files, update it's memory

and we should just have this thing in the background constantly be updated

so that again, we can always go back to a previous checkpoint.

And this is a really good kind of initial setup.

Now for the agents.

We can start actually doing some more serious work okay.

So I think it's good now.

Yeah. So let's save this preference.

Now let's move on to some stuff that is a little bit more interesting with the agent.

Okay.

So let's have a look at voice image and browser automation.

So first let's just check if voice works.

So I'm just going to go here.

Hey does the voice work here.

Can you just tell me what's going on. And I'm just testing essentially.

Just tell me if you can understand what I'm saying and tell me what I said.

Okay.

I'm just going to send a voice message. I want to see if I can understand that. Okay.

So you can see that it was actually able to transcribe this and it gave me the description.

So if you just want to talk to it

using the voice mode, you can I'm going to say can you just say hello world to me.

Can you send me a voice message

and let's see if it can send me a voice message?

I'm just testing this just to show you that by default, if you did, the settings that I did,

it should automatically have like text to speech and speech to text.

Let's say hello world.

But and it does have that okay.

So it's using kind of a crappy text to speech.

But it does have that and it does work.

Now another thing I'm not going to show,

but you can upload images here if you're using an image enabled model.

Most of the modern ones are meaning they can read images and look at them like GPT, opus, whatever.

So if you want to attach an image here,

multiple images, you can do that and it will be able to understand it.

Now the next thing to set up is browser automation.

So let me show you something.

Hey can you go to tech with Tim Dot net?

Take a screenshot and tell me what's on the page.

This is a prompt you can try with your model as well.

Usually by default the browser automation doesn't work really well.

And it needs some additional configuration.

So let's give it a second

and you can see right away it's saying hey, Chrome isn't installed for the browser tool.

So install the browser runtime first and then retry says the agent browser is not installed.

So essentially what I was getting at with this prompt is that when you first

ask it to do a web based task, usually it fails.

If it's failing what you can tell it to do is install the agent browser.

So it's doing that automatically because it just figured out it needs to do that.

But if you want to do a web based task, first thing you need to tell it is, hey, try to do this.

If it fails and it doesn't correct itself automatically, you just tell it.

Install agent browser and if it installs the agent browser now, it should be able to do

basic web automation tasks for you, which is obviously a really important skill to have.

You know, what do you call it for the model for the agent?

Okay.

So you can see that it was able to do this.

It says, you know, here's what's on the page.

Here's the here's action main cards, topic buttons about section, community, whatever.

And it was able to deduce that based on what it found in the screenshot.

All right. So that worked right. Give me the description.

Now what I want to do is show you a really great integration that you should add

to, Hermes agent to connect to things like your Gmail,

your Google Calendar, and pretty much any personal accounts that you have.

Now, this is a tool called Compose You.

I'm not sponsored by them.

I just use them all the time.

I have worked with them in the past, but this is not a sponsored integration.

They didn't pay me to say this.

It is free to use.

You get, I think, 20,000 tool calls per month.

And effectively what Composio allows you to do is just have one connection to your agent.

So you just say, hey, agent connected Composio.

And then from compose you, you can connect handling to all of your individual accounts.

So if you want to connect to Google, if you want to connect to GitHub,

if you want to connect to Zapier, I don't.

Whatever. I'm just naming random things.

You can do almost all the connections

in this nice compose your dashboard and then you just connect compose.

You go to your agent and then it can start actually taking actions for you.

I'm going to show you what I mean, but just go to the compose your website.

Of course there's a fly flying around here.

You want to go to the dashboard, decompose your dev,

that's where you want to be on and just make a new free account from here.

What we're going to do is we're going to go to Connect Apps.

From Connect Apps.

You'll see there's literally thousands of different options.

And this is where you can just go and start

connecting to anything that you would want your Hermes agent to be able to touch.

So what makes Hermes agent useful is when I can actually do something right,

when it can draft an email, when it can read through, when it can do automatic accounting for you,

when it can connect to your notion, when it can go to Airtable, when it can search on YouTube, whatever.

So you need to ask yourself, first of all, again, what am I comfortable giving Hermes agent access to?

Right.

And then start connecting it so that it can actually start doing something.

So I'm going to connect Gmail okay.

And I'll just connect to, I don't know, one of my burner Gmail accounts.

Okay.

And when you connect it's going to ask you for a bunch of permissions.

It's safe to give all the permissions in compose yo, because you can modify

the tools that are accessible later and what it will do here.

If we just give it a second, is let's wait, connect.

Okay.

And you can see that

we've now connected to this account and there's a list now all of these available tools.

Okay.

And the nice thing about compose is you can connect to multiple accounts at the same time.

So if I want to connect to another account, you just press done on this thing because it says

you need to like share it with someone or whatever, you can have multiple different connections.

So you can have like five different email accounts, different Google drives or whatever.

So I've connected Gmail, I'll just connect one more.

So I'll just connect Google Calendar with the same account,

and you'll see a list of all of the tools that are going to be exposed here.

When we connect compose to our agent one check.

Okay, so I've connected to the Google calendar.

And again like you're going to spend some time making all these connections here.

We just need to do them one time.

Next we're going to go to install.

And when we go to install here what we're going to be looking for here is open call.

So just find where it says open claw.

And you can see that it kind of has these instructions.

So I'm just going to copy these instructions here.

And I'm just going to paste them to my Hermes agent okay.

And I'm going to say can you run these instructions and help me connect to compose you.

Oh please.

Let's it. Okay.

So I'm just taking the open claw instructions

and pasting it to Hermes agent because it's effectively kind of the same thing.

And it should now do what's required and let me log in to compose yarn.

Okay.

So now it's telling me to open this URL in my browser.

So let's go here.

And oh actually never mind.

It's now saying it's going to use no browser wrote okay, so wait a second and let's do this okay.

Just open this URL in your browser to approve.

Let's open it.

And we're going to go on authorize.

And let's give it a second.

And now I think it should be good.

So let's go back to here and we'll say

just approved it and hopefully we're signed in.

Let's see okay.

So it says compose yo is connected.

I'm signed in with my email.

And now the compose is connected effectively.

What I will have access to is any of the connected apps that I've connected through compose. You.

I don't know why

it's failing to load them, but anyways, point is, I can just connect them from the dashboard.

Now it's connected to my agent and now anything that I connected I can ask compose

yo to execute so I can say, hey, just tell me what the two last emails were in my inbox,

and now it's going to have access to my email because I've connected it from compose. Yo.

So this was kind of the critical step is like, how do we actually connected to real world tasks

that are useful?

And now it's going to be able to actually go and do this.

Now at the beginning it may use the wrong skills because you see, it's

trying to use this Google Workspace skill, when really it should use a compose yo skill.

So I'm going to kind of train it and teach it and say, hey, whenever you need to look up an email,

use compose yo.

Okay.

Now for some reason it's saying that Gmail is not linked actually.

Okay.

But anyways, it's giving me a link that I can press to link Gmail.

So let's just press this and see if we get this to work. Okay.

And there we go.

Now works and it's connected.

I don't know why

the initial connection didn't work, but you'll notice that it just gives you these links.

And then you can just off it.

And it's all good to go because anyways, it gives me the two emails.

I'm just going to blur them because, they are sensitive actually.

And I shouldn't have shared them the same thing with calendar.

I could tell it, hey, go in, create a new calendar event for me.

Hey, go to slack and read my new messages.

Whatever.

So what you're going to do is spend the time, figure the tools you're going to use,

connect them to this agent.

And now we can start setting up some actually really useful workflows.

Let's get into that.

So because I do a lot of stuff with email, the first workflow that we'll just get into again

is just to give you some inspiration and ideas is a daily email triage.

So what I'm going to do is just write a long prompt.

And I'm effectively just going to tell it, hey, every day at 7 a.m., I want you to read

through my emails from the night before and just give me a quick summary of any action items.

Okay?

So I'm going to say, hey, I want to create a email triage scheduled task.

I want this to run every day at 7 a.m., and I want you to go to my email,

and I want you to look at all

the emails from the past 24 hours and just tell me anything that's important.

There's an action item that I need to know about

and send it in a really concise, neat summary to me again every morning at 7 a.m..

Going to send it in this chat here.

Boom.

We press enter and now we're just going to create this workflow right.

And it's going to take a second and set it up and you know do a test and all this kind of stuff.

And now every morning when it sends us this message, if we want to revise it or adjust it,

we just tell the agent and it will start learning how we like this message to be sent.

Building a skill, updating its own skill, etc.

that's what this agent does.

It builds its own skills based on how we use it.

So the more you use it, the better it gets.

So that's the workflow one.

And again it's just a sample prompt like it's super simple to create these different things.

Now let's move to another okay. Sorry.

So I'm just going back because the previous task wrapped up.

And I just want to show you that it's finished now.

So that's what it's going to do at 7 a.m..

And if I want to actually see this I can go to GitHub.

I can refresh my repo here.

If I press on commits, I'll be able to actually see a history of what's gone up on here.

So you can see that we backed up the daily email trio scan scans, and let's go back here.

And now if I go to, cron, for example, and I go to jobs dot Json,

you can see we have the prompt that it's going to run every single day at 9 a.m..

Right.

And this is the cron task that's been created. Cool.

So here's another example of something that we might want to do

every weekend.

That's on Sunday at 12 a.m..

I want you to run a security audit or wait to check the virtual private server, see if there's

any unusual activity, and send me a report with any suggestions on ways that we can improve it.

Boom. So it's another one security task. Set that up.

Now I'm just going to give you a few other ideas for things to do here rather than prompting

all of them out, because really, now it's up to you to make this useful for your task.

Typically, what people like to start with is a morning brief or digest,

where you connect to a ton of different sources.

It pulls the info and then gives you like the import to action items.

Another thing that people like to do is have it checking their calendar, seeing if they can

reorganize it, or automatically scheduling blocks

to work out or eat or focus or something like that.

Another thing that's quite common is having a do like automated accounting.

So you can say, hey, anytime I receive a new email, I want you to check if this is a receipt,

if it is uploaded to Google Drive and put it in a receipts folder for me.

Right. That's something that's super common.

The more integrations you give it, the better it gets.

Another thing is like a daily wrap up.

At the end of each day, I want you to tell me what you achieved, what you did,

what was completed, and how we're going to improve for the next day.

Xyzzy right.

And again, the more you chat with this, the more you ask it to do stuff, the better it gets.

So this can automate a lot of things for you, but you have to know what it is

that you want to automate.

For most people, there's not that many things they really need an AI agent to be doing.

So that's why I'm sharing with you.

These would seem like simple examples, but can actually be super useful.

Automated accounting.

That's something I have a Hermes agent doing, which is really good.

It just checks my email every day and picks up all the receipts and stores them.

Right? Maybe even a habit tracker, for example.

I'm trying to learn Indonesian right now.

I'm in Bali at the moment, recording.

Hey, every day I want you to teach me five new Indonesian words.

I want it in this format.

And then, you know, quiz me on them, ask me them and ask me to say it.

Invoice to you, whatever. Right.

So you can set this up to be really, really unique and creative.

And it's just about getting the right integrations in place.

This is not magic.

This is not going to, you know, make you $1 million overnight, as everyone online will tell you.

But it is extremely useful when set up correctly.

And that's why I went through this full, long process of giving you, hopefully, a really solid base,

where now you have a Hermes agent

that really can be configured just by you creatively asking it to do something

it doesn't require, you know, rocket science or mission control over all of this.

It's just coming up with real legitimate use cases and then prompting the model

and working on it over time.

So anyways, guys, that's going to wrap up this video.

That was a lot of stuff.

I hope you guys found value here.

If you did make sure leave a like subscribe and I will see you in the next one.