HomeVideos

How to USE the #1 Personal AI Assistant (Step-by-Step Setup)

Now Playing

How to USE the #1 Personal AI Assistant (Step-by-Step Setup)

Transcript

551 segments

0:00

What I'm seeing a lot with founders,

0:01

they're using AI to do a lot of dumb

0:03

things really fast.

0:09

How can we use AI to do the things that

0:11

are the normal functions of the business

0:12

way better, way faster?

0:14

>> Everybody will have an AI that is an

0:16

assistant.

0:16

>> It really feels like my chat GBT gets to

0:18

know me.

0:21

>> Hermes agent is one of the most powerful

0:24

AI tools in the world right now.

0:28

Most people are using AI the completely

0:31

wrong way. They're using it for building

0:32

websites and voice agents and things

0:34

that they actually don't need. The

0:36

entire point of AI is to make you more

0:38

productive. But funnily enough, most

0:40

people use it in a way that makes them

0:41

unproductive. Just because AI can do

0:44

something doesn't mean that you should

0:45

do it. And if you're looking for the

0:46

next flashy thing to hit your dopamine,

0:48

then this video isn't for you and you

0:50

might as well click off. But if you want

0:51

to use AI to actually become productive,

0:54

then this will blow your mind. In this

0:56

video, I'll show you how you can set up

0:57

your own personal assistant that you

0:59

would usually pay $4 to $5,000 a month

1:02

for. I'll show you how you can do it in

1:03

10 minutes without needing to be

1:05

technical, and I'll skip all of the

1:07

fluff that you actually don't need.

1:08

Let's not waste any more time. Let's get

1:10

into it. Let's start with the most

1:11

important thing, which is the right

1:13

setup. Most guides on YouTube show you

1:15

complicated setups that requires you to

1:17

be technical, but this video will be the

1:19

exact opposite of that. Our personal

1:20

assistant needs a computer that's turned

1:22

on at all times, and that's what I've

1:24

set up right here. And I'll show you how

1:25

to do the same in 10 minutes. And it's

1:27

literally a one-click install. And the

1:29

entire setup will be done with our words

1:32

like chatting to chat GBT. We can do

1:34

that to host our virtual computer. And

1:36

to set it up with only chatting, we're

1:37

going to use this platform right here.

1:39

That's called abacus.ai. But we're not

1:41

going to use the regular chat. We're

1:42

going to set up our own computer. We do

1:44

that by just going top left into

1:46

supercomput. And don't worry, there's a

1:48

terminal here on the right hand side,

1:49

but we won't really be using that. Now

1:51

we have a virtual computer that lives

1:53

somewhere else, which means we can have

1:54

it so it's always on. That becomes very

1:57

crucial later when we set up the

1:58

personal assistant. When you're on this

2:00

page right here, then it's literally a

2:02

10-minute setup and it's a one-click

2:04

install. You've probably heard about

2:05

OpenClaw, maybe also Hermes agents.

2:07

Right now, OpenClaw is kind of dead and

2:09

the best personal assistant you can have

2:10

right now in my opinion is Hermes

2:12

agents. So, we can just click on this

2:14

and you can see it says set up Hermes

2:15

agents and right off the bat we're going

2:17

to have AI credits. We're going to have

2:18

web search and I'll also show you later

2:20

in this video how we can add our own

2:22

OpenAI account to have really cheap AI

2:24

credits. That is crucial when you want

2:26

to run this. Another thing that's

2:27

important is that we want to turn on

2:29

this right here, which is to have our

2:30

computer always on. We want to be able

2:33

to text our personal assistant in the

2:34

middle of the night and still have it

2:36

respond. So, I'm going to turn enable

2:37

always on. Click continue and then click

2:39

next. And now you can see it says

2:40

setting up the agent. It takes 30 to 60

2:43

seconds. And now we are done. Usually

2:45

setting up a virtual machine and

2:46

installing Hermes would take you maybe

2:48

30 minutes to an hour. We did it in one

2:50

click. But if you get this next thing

2:52

wrong, then your credits are going to

2:53

cost you 10 times as much. We want

2:56

access to the best models possible

2:58

without having to pay a ton for credits.

3:00

Let me show you the best possible setup.

3:02

Click on this little Hermes icon right

3:04

here. And what you'll be able to see is

3:05

that you can either configure it with

3:07

Claude or with the codec. What's worth

3:09

knowing, however, is that Claude

3:11

recently changed it. So you cannot use

3:12

your Claude subscription. you need to

3:14

pay for an API key which means cla is no

3:16

longer valid. Instead, we're going to

3:18

use the codec cli. So, click on that.

3:20

And now you can see it says sign into

3:21

openi codeex. It's going to give us a

3:23

URL which is this one right here. So,

3:25

we're going to copy this, place it into

3:27

our browser where we are logged into

3:29

chat gbt. Going to log in with this

3:31

account right here. If it's not enabled

3:32

on your chatbt account already, you need

3:34

to click right here. It is going to open

3:36

the settings panel. Scroll all the way

3:38

down and turn on enable device code

3:40

authorization just like this. And now go

3:42

to this link again and it should work.

3:44

Click continue. And now you can see we

3:45

need to insert the code which is this

3:47

code that we've gotten right here. So

3:49

I'm going to copy this, paste it in,

3:51

continue. And there we go. Signed into

3:53

Codeex. And if we go back, you can see

3:55

it says added OpenI Codex or

3:57

credentials. Just like that. We can then

3:59

click next. And now it's setting up the

4:00

codeex CLI. Basically the way this works

4:03

is that it's now going to use our chat

4:04

GBT plus subscription and all the

4:06

credits that is connected to that. and

4:08

you have a lot of credits inside of your

4:10

regular chatbt subscription. So if you

4:12

have one already, I'll highly recommend

4:14

that you use the codec cli. Now we can

4:16

enter full screen right here, which is

4:18

going to open up the full screen for the

4:20

MS agent. And the next thing we want to

4:21

do is of course switch the model to use

4:23

our oorthth instead. So what I'm going

4:25

to do is that I'm going to say please

4:26

use the model GPT 5.6 and change the

4:30

model so it uses our codeex

4:32

authentication CLI. Code type is going

4:34

to improve that prompt. Write it in like

4:36

this. We can now hit enter. You can now

4:37

see that the model it's using is GBT

4:39

5.5. And if I go to this section right

4:41

here where it says model, I can choose

4:43

whatever GBT model that I want to use.

4:45

Like if I wanted to do something

4:46

extensive, I could use the newest 5.6

4:48

soul model and switch to that if I

4:50

wanted to. And you can see this is using

4:52

the OpenAI codeex that we authenticated

4:54

with, which gives us a lot of credits.

4:56

This is the best way that you can set up

4:58

your personal assistant without having

4:59

to pay $200 a month for claw. Now we're

5:02

ready to build out our personal agent so

5:03

it can actually do stuff. Very excited

5:05

about this. A mistake that many

5:07

beginners make is that they create an

5:08

agent but it lives on their computer

5:10

which also means that they can only use

5:12

it when they are sitting at their

5:13

computer. But the idea of having a

5:15

personal AI assistant is that it runs

5:17

all the time and that you can send it an

5:20

idea in the middle of the night from

5:21

your phone and it'll pick it up and

5:23

actually do something. But the entire

5:25

idea of a personal AI assistant is that

5:27

you can send it a message in the middle

5:28

of the night from your phone with an

5:30

idea that you have. So we need to do a

5:32

bit more setup. But don't worry, it is

5:33

really simple. We can just say please

5:35

enable Telegram so I can contact my

5:38

agent from Telegram. I'm just going to

5:40

dictate this instead. And then we hit

5:41

enter. And because we already have a

5:43

pre-built Telegram integration inside of

5:46

the MS agent, we can now enable Telegram

5:48

very very quickly and run it all from

5:50

our phone. It's going to ask you

5:51

permission to run this command. So we

5:53

can use our arrow keys to go down and

5:55

say allow in this session and then hit

5:57

enter. And now I can see it says to

5:59

enable telegram I need your telegram bot

6:01

token from buttf father and the telegram

6:03

user ID. I'll of course show you how to

6:06

get that from your phone on telegram

6:07

search up botfather and this is the

6:09

account that's used for creating

6:11

telegram bots. Now write slash and then

6:14

new butt like this. It now says all

6:17

right a new butt. How we going to call

6:18

it? Please choose a name for your butt.

6:20

I'm going to call it wyb butt and then

6:22

hit enter. You can see we need it to be

6:24

low capital like wordyb butt like this.

6:27

That name was Ting. We can try it like

6:28

this also. Ting. Maybe just do it like

6:30

this. Albot assistant bot. Everything is

6:32

Ting. Do it in just one word. There we

6:34

go. It now says done. Congratulations on

6:36

your new bot. You'll find it. Add. And

6:37

then you have the link. And now we need

6:39

this token right here, which is token.

6:41

So only we can access it. So we're going

6:43

to copy this token. And then inside of

6:46

the Telegram, we can paste it in here

6:48

and say this is the token. And then just

6:51

hit enter. is now going to use their

6:52

secret token and set it up so only we

6:55

can contact our agent which is very

6:57

important for security. Last thing you

6:58

want to do is give it your telegram ID

7:00

and when it's done it's going to set up

7:01

Telegram for you. And the last thing you

7:03

need to do is that you just need to go

7:04

to your bot and click start like this.

7:07

Next thing you want to do is that you

7:08

want to search up user info bot and

7:10

click this one right here and then just

7:12

click start and that's going to give you

7:14

the numeric user ID. And that is all we

7:16

need to do. The agent is now going to

7:18

set all of this up for itself. So we can

7:20

very soon contact it from Telegram.

7:22

There we go. Telegram is now enabled and

7:24

verified. And now I received a

7:26

notification. And there we go. MS

7:28

Telegram gateway is connected. You can

7:29

now message this bot to reach your

7:31

agent. We can write, hey, can you hear

7:33

me? And you can see we receive a

7:34

response. Yes, I can hear you. Now we

7:37

can contact our personal assistant

7:38

anytime from Telegram. Isn't that cool?

7:41

This becomes crucial for the later parts

7:42

of this setup. This next trick will

7:44

literally save you hours because you

7:47

don't only want to be texting your voice

7:49

agent. You also want to be able to send

7:51

voice messages with long ideas that you

7:53

have and you want the agent to be able

7:54

to understand all of that. You could

7:56

either have something like a local

7:58

whisper model. However, for the most

7:59

accurate and fast results, I recommend

8:01

using something like 11 Labs. Just go to

8:03

the 11 API, click API keys, create a new

8:06

key, and I'm going to call this MS

8:08

agent. And then I'm going to unrestrict

8:10

the key and create an API key like this.

8:13

Now we can copy this API key. Go back

8:15

into our Hermes agent. Go into keys and

8:19

then I'll just search for the 11labs API

8:21

key and we can set it. Paste this in and

8:23

click save key. There we go. Now we want

8:25

to create a new chat session just by

8:27

clicking new chat. And then I'm going to

8:29

say I just added a 11 Labs API key. I

8:32

want to enable voice messages for

8:34

Telegram and I want to use the Scribe

8:36

API from 11 Labs to transcribe the

8:39

messages so you also understand those.

8:41

Can you please set that up? Thank you.

8:43

That's going to improve the prompt.

8:45

Write it in like this. And now we can

8:47

hit enter. And the thing that's so cool

8:48

about these personal AI assistance is

8:50

that I can actually just tell it to

8:51

improve itself and it does it. One thing

8:53

that's very important is that when you

8:55

add your API keys is that you do it

8:57

through the keys tab right here. You

8:58

never want to give your API key directly

9:00

to the AI. It's much more safe doing it

9:03

the way that I showed you. It asks us

9:04

for approval. So, we can say allow for

9:07

this session and hit enter. And it's

9:08

going to build out this entire

9:10

integration itself. Awesome. This is now

9:12

fixed. I can see that it just sent me a

9:14

message right here where they actually

9:16

sent me a voice message. Voice messaging

9:18

and 11 Labs Scribe transcription are now

9:21

enabled.

9:22

>> That's cool and all. However, I want to

9:24

receive text and then be able to send

9:26

the voice messages, right? So, I'll just

9:28

say, "Please enable it so I can send

9:30

voice messages." And you understand? But

9:32

you respond back with regular text.

9:35

Paste that in. This should just be an

9:37

easy setting change. And then we should

9:38

be good. There we go. This is now set

9:40

up. I can now send a voice message.

9:42

Hello, can you hear what I'm saying?

9:44

Hello, can you hear what I'm saying? It

9:45

says it back. But we wanted to be able

9:47

to respond, of course. And there we have

9:48

it. Yes, I can hear your message

9:50

clearly. Hello, can you hear what I'm

9:51

saying? It's a good idea. We could also

9:53

disable it if we don't want to. So we

9:54

can actually hear like we can see what

9:56

it wrote. But I actually kind of like it

9:58

because if for example the transcription

10:00

gets something wrong I have time to

10:02

change it. If it doesn't do this then I

10:03

don't know if it gets something wrong.

10:04

Perfect. Now Telegram is enabled as well

10:06

and I can close down this tab right here

10:08

because this is running at all times and

10:10

we have an agent that's live that I can

10:12

message from my phone at any time. But

10:14

this is where it gets really interesting

10:15

because now we have the base set up.

10:16

It's time to actually give our AI

10:18

capabilities so it can do stuff for us.

10:20

Let's get into it. Now comes the most

10:22

important part of this entire setup,

10:24

which is setting up our agent with the

10:27

things that we actually want to automate

10:28

and the things we actually wanted to do.

10:30

And this is the exact way that you

10:32

actually become productive using AI

10:34

instead of unproductive. And I'll show

10:36

you a tool, one tool that will allow you

10:38

to integrate into absolutely everything,

10:41

all the tools that you use. First thing

10:42

we're going to do is that we're going to

10:43

go inside of keys and then we're going

10:46

to scroll all the way to the bottom

10:47

where it says custom keys. I'm going to

10:49

write composio and then underscore api

10:52

and then we can click add. And here we

10:54

need to insert a value. Now go to

10:56

composio.dev. And what composio

10:58

basically is is that it's a way to give

10:59

your AI access to a bunch of different

11:01

tools. And we can get started completely

11:03

for free. When you've created the user,

11:05

then you go inside of API keys on the

11:06

left hand side. You click create API

11:08

key. And I'm going to call this MS

11:10

agent. I'm going to allow it to write

11:12

all. And then click create. And this

11:14

gives us an API key right here that we

11:16

can copy and then go back into our

11:19

Hermes agent. Paste this in and click

11:21

save. There you go. Compose your API key

11:24

is now saved. Now we go inside of chat.

11:26

I'm going to create a new session. And

11:28

then we're going to integrate into our

11:30

first tool. All I have to do is

11:31

literally just say I want to give my

11:34

agent access to Gmail as well. And I

11:37

want to do that through the Composio

11:39

API. I've added an API key that we can

11:41

use. set it up and give me a link so I

11:44

can authenticate with my Gmail and so we

11:46

can reply to all of my emails with the

11:49

Hermes agent. This will now improve the

11:51

prompt. I'm going to codeype it. It's

11:53

going to write it out and then we're

11:54

going to hit enter. While it's doing

11:56

that, let me show you a bit about what

11:57

we can do with Composio. Composio

11:59

basically allows you to integrate agents

12:01

into a bunch of apps like Gmail, Google

12:04

Calendar, Slack, GitHub. There's a bunch

12:06

of integrations that you can integrate

12:08

with. Basically, all of the tools that

12:10

you use and they have around a thousand

12:11

apps inside of here. So, by using

12:13

Composio, we only need to have one API

12:15

key and then we can give our agent

12:17

access to all of the tools that it

12:19

needs. You can see it's asking for

12:20

approval right here. We're going to say

12:22

allow for this session. And it's now

12:23

going to set everything up for us and

12:25

help us connect our Gmail via the

12:27

Composio API. There we go. Now, it has

12:29

set up Composio. So, what we need to do

12:31

is that we need to click on this link

12:32

that is it has given us right here and

12:34

click continue. We can check the link

12:35

that it is composure. It says that it

12:37

blocked the popup. So I'm just going to

12:39

click allow pop-ups and redirects. Done.

12:41

Can click the link again. Click open.

12:43

Now it opens up this link right here

12:44

where we can authenticate with our

12:46

Google account. So I can just click

12:47

continue. And then I can give it access

12:49

to everything that it needs. And click

12:51

continue again. And Composio allows it

12:53

to give it access in a secure way. It

12:55

basically manages the connection for us.

12:58

So we can close this down. And I can go

13:00

back. And now I can say I just connected

13:03

my Gmail. Can you please check if the

13:05

connection works like it should and just

13:06

dictate that in. Hit enter. And now

13:09

let's see if it has the right access. It

13:11

should have because we just logged in.

13:13

And there we go. You can see it is now

13:15

connected successfully. So now we can

13:17

create a new chat. And I'm going to

13:18

change the model to be a little faster.

13:20

And now we can say, can you please check

13:22

the latest emails that I've gotten? Hit

13:24

enter. And because we have connected it

13:26

to our Gmail, it will now be able to do

13:28

just that. You can see it's using the

13:29

Composio MCP to fetch our emails. And

13:32

now I can show the latest emails that we

13:34

just received. And now we can just

13:35

connect all of the tools that we are

13:37

using via Composio. So I can say I also

13:39

want to add my Google calendar as well

13:42

also via the Composio setup. Can we

13:44

please fix that too? Paste that in. Hit

13:46

enter. And it's now going to create the

13:47

Composio connection as well. There we

13:49

go. Just like before, it's going to give

13:51

us this link that we can open and

13:53

authenticate our Google calendar. So

13:55

just like before, I'm going to

13:55

authenticate and click continue. And

13:57

there we go. Successfully connected. And

13:59

if we go back, we can now say, I just

14:01

connected it. Can you check that the

14:02

connection is still active? Hit enter.

14:04

And it should come back. And I'll say

14:06

that it can see the connection. And

14:08

there we go. Now we have also connected

14:10

our Google calendar. And this is where

14:13

it becomes magical. We can now create a

14:15

new chat. And then we can say, every

14:17

single morning at 9:00 a.m., I want you

14:19

to check my calendar and my emails and

14:22

give me a short daily briefing of

14:24

exactly what I'm doing today. And send

14:26

that to me on Telegram. Please use the

14:28

composio connection for this. Thank you.

14:30

And then we can hit enter. And the great

14:32

thing about Hermes agents is that it can

14:34

create jobs for itself. This computer

14:36

that we on right now never turns off,

14:39

which means that it's always on. What's

14:41

going to happen now is that the Hermes

14:42

agent can give a job to itself at 9:00

14:45

a.m. every single day to finish some

14:47

task. And this becomes extremely useful

14:49

for every single thing that you want to

14:51

automate. Just by knowing how to do

14:52

this, you can connect the exact tools

14:54

that you want through Composio. You can

14:56

tell the agent that you want to do this

14:58

and this and this and it's not going to

14:59

build it out for you. I hope you start

15:01

to see the power of this. Just by

15:03

knowing this, just by using natural

15:04

language, you can automate anything that

15:06

you want. And there we have it. You can

15:08

now see it says scheduled successfully

15:10

every day at 9:00 a.m. It's going to

15:12

give us this daily calendar and email

15:14

briefing. So to test it, we can

15:15

literally just ask it, please run a test

15:17

of the daily brief for today. Hit enter.

15:20

And now hopefully we should receive a

15:21

telegram message with the daily briefing

15:24

of today. And there we go. on my phone.

15:26

We just received our daily calendar and

15:28

email briefing. I can see what things I

15:30

have to do today. I can see the unread

15:32

emails. Absolutely all of that. How cool

15:35

is that? I hope you see the power of

15:36

this because when you know how to do

15:38

these two things, one connect any tool

15:40

via Composio and two, how to set up a

15:43

crown job like we just did right here,

15:45

which is basically just a schedule task,

15:47

then you can automate anything in your

15:48

life. But what should you actually

15:50

automate? Most people when they get

15:51

access to tools like claw code and like

15:54

the AI models they start building stuff

15:55

that they otherwise wouldn't need like

15:57

they start building a new website or

15:59

they build some flashy voice agent that

16:01

can do some random thing. That is not

16:03

how you should use AI and that actually

16:05

makes AI unproductive for you instead of

16:08

productive. If you're running a business

16:09

right now and let's just take an

16:10

example, you are selling pens. You don't

16:13

want to automate things that don't help

16:14

you sell more pens. You want to take a

16:16

look at all the things that are not

16:18

selling pens and that's the things that

16:19

you want to automate. So you checking

16:21

your emails in the morning, that's

16:22

something you can automate. You checking

16:24

your calendar, something you can

16:25

automate. Your invoicing process, your

16:27

onboarding process with this setup you

16:29

have right here that is running 24/7.

16:31

You can literally just talk to your

16:32

phone and say, "Please onboard this

16:33

client. They are inside of our CRM. Send

16:36

them the contract. Thank you." And then

16:38

that entire process that would usually

16:40

take you 10 to 15 minutes opening up the

16:42

laptop doing everything done the right

16:44

way every single time. That is the power

16:46

of learning how to set this up. But now

16:48

it's your turn. Don't let this video be

16:49

another one of the videos that you watch

16:51

on YouTube but never actually take

16:52

action on. Go set up your own personal

16:54

agent and automate the things that

16:55

actually moves the needle in your

16:57

business. If you want more tips and

16:58

tricks about how you can build a

16:59

business with AI, then you can check out

17:01

our community. There's $165,000 members

17:04

inside of here right now and you can

17:06

start on a completely free trial.

17:07

Charlie just closed a $3,500 a month AI

17:10

automation client from the things he

17:12

learned inside of their community.

17:13

Inside of here, you also have all my

17:15

claw skills and training on how you

17:17

build with AI. I'll leave a link for

17:18

that right below in the description.

17:20

Thank you so much for watching. I hope

17:21

you enjoyed the video. Have a wonderful

17:23

rest of your day. The elbow out.

Interactive Summary

This video provides a step-by-step guide on creating a 24/7 personal AI assistant using Hermes agents and Abacus.ai. The instructor demonstrates how to set up the environment, integrate it with Telegram for remote control, enable voice message processing via 11 Labs, and use Composio to connect the agent to productivity tools like Gmail and Google Calendar. The video emphasizes using AI for meaningful business automation—such as managing emails and scheduling—rather than building unnecessary flashy features, ultimately teaching viewers how to automate daily tasks to increase productivity.

Suggested questions

4 ready-made prompts