Stop Using Claude Without an Agentic OS

If you've worked with a second brain or

memory in cloud, you know how powerful

it can be for AI to always be able to

pull in relevant and up-to-date context

around you and your business. But what

makes this even more powerful is to have

a personalized dashboard or command

center on top of it to manage your

intelligence and to automate work. So in

this video, I'll show you the four big

benefits of having this dashboard or

command center set up. I'll show you

what an agentic OS actually is, show you

the three options you have to build one,

and an easy way to set this up for

yourself today. Now you might have heard

these terms like agentic OS or command

center, and there are many other fancy

terms being thrown around, but all it

really means is to have a personalized

dashboard like this set up with all the

relevant and up-to-date context in your

business in one centralized place. Now

there are four main benefits to having

this set up, and the first one is of

course you can have a completely

personalized intelligence or UI

dashboard for yourself with access to

live data from all of your softwares and

data from your second brain if you have

one. And it can be molded into the a

custom UI that presents you with the

most relevant information for you. For

example, in my research tab, I can

instantly see the latest information on

AI and tropic updates, YouTube trends. I

can see the top Reddit AI discussions. I

can see competitor activity. In my comms

tab, I can see unreplied messages across

all my communication channels like

LinkedIn DMs, my Circle AI community,

YouTube, and my email, and all other

relevant and up-to-date intelligence I

need for my day-to-day work. And of

course these dashboards can be

customized for each of our team members.

For example, for one of my sales reps,

his dashboard is customized to his most

important day-to-day work in sales. And

we even have a company-wide dashboard

for general intelligence and analytics.

But besides that, this command center

also allows you to take actions with AI

right from that dashboard. I can, for

example, directly use skills by clicking

buttons to, for example, repurpose my

YouTube video into a LinkedIn post.

Because it's connected to my softwares

through MCPs, it can even take actions

right inside of those softwares. For

example, I can respond directly on

LinkedIn here. I can use a skill to

draft a reply and then actually send it

out through LinkedIn right from my

dashboard. I can also spin up an AI

conversation anytime inside of this

dashboard. And these agents, of course,

have access to all of the context behind

this dashboard. I can also set up

specialized long-running agents like my

YouTube research agent here and run

them. And I even have a widget here

where I can chat with AI that has

context on the dashboard I'm looking at

right now. And these AI capabilities are

not limited to one model. Because a

command center like this becomes model

agnostic. As you can see here, I can use

Claude, but I can also use Codex or I

can integrate it with any other AI

provider that I want. Any of these

models still has access to all of the

context, of course, behind this

dashboard. And lastly, of course,

because it lives on a live URL, we can

really easily share it with team members

or share this with potential AI agency

clients that we set this up for. Now,

because of this combination of things,

you can imagine this can really become

your single command center or operating

system for doing work instead of being

in the cloud desktop or in the terminal

or hip-hopping between many of your

different softwares. Now, we actually

have three options to set this up.

Besides this live URL, I've also set one

up inside of Obsidian and another one

which is a live artifact inside of the

cloud desktop.

Later in this video, I'll give you the

advantages and disadvantages of each of

these setups, give you my

recommendation, and then show you for

each one a simple way to set it up. But

before that, let me quickly go over how

this actually works because it might

look very complicated, but it's actually

not that complicated. And the simplest

way to understand how it works is by

understanding the layers of an agentic

OS or an AI operating system, whatever

you want to call it. Now, first, of

course, we have our L&M layer with our

AI models like Claude, ChatGPT, or

Gemini. Then we have the memory or the

context layer, which usually live in the

folders on our computer with markdown

files that contain context on us, our

business, and even up-to-date context on

everything that's changing and happening

across our business. And with this

context and memory, these LLMs can of

course give us far more relevant and

better outputs, but agents with like

Claude and ChatGPT can of course also

save to this memory, so we get

persistent context and memory across

different chats, different AI providers,

and across different team members. And

this context or memory is of course just

a folder on your computer with markdown

files, text files, or you can be using

Obsidian for this memory or context

layer, which is just a tool or an app

that helps you visualize a folder with a

lot of files in a better way, which is

also known as the second brain. Now, if

you have no idea what this is yet, or

you haven't set this up yourself, I've

two full videos covering how to set up a

second brain. So, if you haven't yet,

make sure to check that video in the

link in the description below, but you

should be able to follow this video,

too.

Then next, we have the capabilities

layer with features like scheduled

tasks, routines, skills, and loops that

allow these agents to actually do tasks

and work for us autonomously. And then

we have the connector and MCP layer that

allows these agents to actually get data

and take actions in our softwares. And

because we have these capabilities like

for example scheduled tasks, we can for

example pull up-to-date context from our

softwares or the internet, like meeting

transcripts from our email inbox,

YouTube data,

recent news, and feed it that

intelligence and that data directly into

that context layer or memory layer. And

it's how we provide that memory layer

with up-to-date and relevant

intelligence for us. But there's still

one layer missing, because if you just

use Claude, for example, in the terminal

with Claude code, we need to prompt an

AI model each time to actually to access

to that up-to-date context and

intelligence from the memory layer. So,

what's missing is really that interface

layer, a place where we can actually see

and act on all of this data and manage

the infrastructure in a better way,

which is really the essential thing for

AI to become our main operating system

for work. And that's why, of course,

these AI providers are investing a lot

in developing their desktop apps like

the Cloud desktop or OpenAI Codex or

Google Antigravity. But, the downside,

of course, of using these AI providers

is that they're not personalized and

they're one-size-fits-all and they don't

actually allow us to visualize this

context or memory layer very well. So,

if you want to get a clear insight on

daily intelligence or my analytics, uh

my to-dos, or my comms, I'd either have

to set up different schedule tasks, uh

like here, and go hip hop between them,

look at the markdown files it generated

for each day, usually, of course,

generated in text file, so harder to

digest. And if you get a folder with

hundreds or thousands, even, of these

context files, it just becomes hard to

digest, of course. And even if you use a

tool like Obsidian to visualize some of

these markdown files a bit better, for

example, here, a market research brief,

they're still laid out as long text

files, uh which is just not a great way

for humans to consume data. So, you can

see why setting up a custom or a

personalized dashboard can be a much

more practical. Also, you might have

seen or heard about these tools like uh

Hermes agent or Hermit AI.

These tools basically have their own

infrastructure or framework that

incorporate all of these layers, but

they've added their specific UI instead

of the Cloud desktop or the Codex UI.

But, I recommend setting up one of these

custom dashboards because the biggest

value of a command center like that, in

my opinion, is to have that personalized

UI or intelligence layer. Because if you

set it up well, you immediately have the

most important context to start your

day, to start taking action, to make

decisions, and to know what to

prioritize. And of course, this can be

set up custom for each of your team

members if you're in a business. Now,

before showing you how to set it up, if

this is all going a little bit too quick

for you, you might just be starting with

this and it might be a little bit

overwhelming. We have a full Attentic OS

setup course in my AI Accelerator that

walks you through all of this

step-by-step together with unlimited

one-on-one live tech help, multiple

weekly Q&As with me and my team, and

resources to make all of this a lot

easier for you. So, if that's

interesting to you, you can check out my

AI Accelerator in the first link in the

description below. Also, if you're a

small business and you want me and my

team to actually set this entire

infrastructure up for you together with

consulting and training for you and your

team, you can also book in a free call

with us in the second link in the

description below. And if you might be a

bigger business and looking for a

long-term AI partner, you can also book

in a free call with my AI agency in the

third link in the description below. So,

how do we set it up? Now, first you want

to define what kind of custom dashboard

you're going to build, and I'll show you

how to set up each one after. Now, the

first option you have is to set up a

live artifact inside of Cloud. The

second option is to set up a custom

dashboard inside of Obsidian. And the

last option is the one that I showed you

at the beginning, which is to set it up

through a custom HTML page that can be

deployed on an actual website. Now, the

advantage of a live artifact is that

it's the easiest to set up and the least

technical. The downside is that you

won't really have an action layer inside

of this dashboard because you can't

actually trigger skills or run agents

directly from this dashboard. You can

also not really take actions in

software. So, if you want to take

actions on this data, you'd need to open

a new task. They're also not shareable

with other people and they can't be used

across teams. You also have some

limitations with the UI. But if you're

just going to use it for yourself and

are mostly interested in that visual

layer, not really the actions, and want

to start simple, and maybe already use

the Cloud Desktop a lot, then I'd

recommend starting with this one. Then

for Obsidian, the upside is that you can

actually add in that action layer, and

you have much more flexibility in the UI

and it becomes shareable. So, you

basically get a dashboard overlay here

on top of your Obsidian. And again, you

can have these buttons like escalate

that actually trigger skills or take

actions in softwares. And you can even

build in a terminal here

to directly interact with Cloud Code or

can also be used for Code X or or Gemini

if you use those. Now, it is a bit more

technical to set up, but if you and

maybe your team already use Obsidian,

this is probably your best option. And

I'll show you in a sec how to do it in

an easy way. And then if you want the

most flexibility, you can use the last

option because you're basically building

a custom app, just like the example I

showed you in the beginning. Now, we can

either build this for ourselves by just

deploying it on a local host, or you can

host it on a website with even password

protection. So, it can be shared with

other people. Now, the big downside with

this setup is that any AI action we

integrate into that app or dashboard,

this will be done through the API. So,

it will be significantly more expensive

to take those AI actions in comparison

to the AI actions, for example, we'd do

on the Obsidian setup. So, that's one

thing to keep in mind. And if you're

setting up or planning to to set up AI

OS systems or dashboards for clients or

other people, I'd either go with option

two or option three depending on if

you're going to use Obsidian for them or

not. Now, I'll show you exactly how to

set up each one, but there's one thing

you want to keep in mind for all of

these is first just focus on building

the interface, the action layer where we

add in the skills or we can use agents

or MCP actions, worry about that later

because 80% of the benefits really come

from that personalized overview or

intelligence. And even at that interface

layer, you want to start really simple.

Start with the

essential things

that you want to see when you start your

day and then build on top of that. And

then once you actually start getting

into the dashboard more and more on a

daily basis, that's where you can start

incorporating actions. Me, for example,

that's where I'm at I'm now opening it

daily to just get a quick glance of

everything that's important for me. And

the more I'm doing it, the more actions

I'm I'm adding in because I I see what's

actually helpful and what not. So, very

much take the minimum viable product

approach. That's my recommendation. So,

first, how do we set up a live artifact

command center? Now, before even this

three-step process, if you haven't

actually set up a second brain yet or a

folder on your computer with some

context, this is really the first step

before even doing this. But, I'm

assuming you already have this. If you

don't, don't worry. Just watch my other

video to get the exact step-by-step on

setting up your context folder or your

second brain. Again, I'll make sure to

link it in the description below. But,

once you have that, there will be three

steps. First, we have to set up an MCP

out of our context folder or our second

brain folder. And this means that we

basically have a connector that can

access the context in our second brain

or in our context folder. Now, why do we

need that? Because these live artifacts,

the way they work is they pull data only

from connectors and MCPs. So, they can

pull data from our softwares.

But, of course, we also want to pull

data from this memory layer, this

context layer. And because that lives on

a folder on our computer, we need to

actually build a connector, an MCP, out

of that folder so the artifact can

actually pull data from it. Because then

every time we reload or open the live

artifact, it will automatically be

updated with the context in our second

brain and our softwares. Now, setting up

this MCP is a bit more technical.

The way you would do it is by building

an MCP of the folder on your computer by

using the MCP builder skill from

Entropic, then deploying it on Railway,

and then creating a connector out of it.

You can do this through Claude code.

But, if you're interested in joining my

accelerator, we list all our skills

including our fault MCP skill, which

basically builds this entire MCP out of

your second brain automatically for you.

So, if you're going to use that skill,

all you need is to make sure that you

have Obsidian set up. It's a free app

that visualizes that context folder on

your computer. Then all you have to do

is go here to settings, download a free

community plugin, which is called Relay

here in browse. You go and

search for Relay.

You can You can just install that one

for free. You can then just import our

scale and then it will walk you through

the entire process of setting up an MCP.

It will set up a railway server and then

give you a URL, which you can just copy,

go to the customize tab, go to the

connectors, and add that as a custom

connector.

You add that link there. You can call it

your second brain. And then all you have

to do is log in with the same account as

um the relay plugin and then you have it

set up. Then once you have set up that

MCP, all we need to do is just prompt

Claude inside of Claude co-work to build

us a live artifact with the dashboard we

want. We can tell Claude something like

build me a live artifact. You can give

your specifications and all the

integrations that you want to add in

this live artifact and how the dashboard

should look. And you can build an

initial version of this command center.

Now again, keep it simple at the start.

I also built a skill that helps you get

to a good initial live artifact

according to your best practices fast,

which is also available together with

all our other skills inside of my AI

accelerator. And then the third step is

to actually start using it and then

iterate and improve based on uh your

using it. Now if you want to set it up

in Obsidian, the nice thing is we don't

actually have to set up an MCP out of

our second brain because it also already

lives inside of Obsidian, which is of

course a local app. Because this command

center basically lives inside of one of

the subfolders, as you can see here. Now

before showing you how you would do

this, again, if you're a bit less

technical and want to make this process

of setting up the Obsidian dashboard

really simple, we also built a skill for

our accelerator members that does this

entire process for you and sets up this

initial dashboard directly for you

inside of Obsidian. All you'd need to do

is import that skill, run the skill. It

will then ask you if you want to set up

a web dashboard or an Obsidian

dashboard. Now in this case, I asked it

to do web dashboard, but you can choose

Obsidian dashboard. It will walk you

through your process, ask you some

questions, which connectors you want to

integrate, etc. And it will then set up

an initial dashboard for you according

to your connectors, etc. Then you can

iterate on that in Claude co-work or in

Cloud Code. The scale you'd have to run

in Cloud Code, but you can do the

iteration process in Co-work or

whatever. All you do is you select your

folder that's connected to your

Obsidian, and you can just prompt Cloud

to adjust the dashboard inside of the

folder to any layout styles that are

more relevant to you. And if you use the

scale, you're already have the terminal

integrated with Cloud Code, so you so

you can actually work directly from

here. Now, if you want to build this out

yourself, what you'll need to do is

first to install some Obsidian community

plugins. You'll need the Custom JS

plugin, the DataView plugin, the Shell

Commands plugin, and the Terminal

plugin. And these you need to actually

be able to build sort of an HTML

dashboard inside of Obsidian, and

actually that actually looks nice. So,

again, you could just go here to the

settings, go to the community plugins,

install those four, like Custom JS,

DataView, Shell Commands, and Terminal.

And once you have those set up, you can

just prompt Cloud in Cloud Co-work or

Cloud Code or even OpenAI if you use

that. Give it access to the folder of

that second pane in Obsidian, and ask it

to create a new folder, which is called

dashboard. And then you can basically

tell it to create your dashboard in your

way. Now, important to also ask it to

integrate your MCPs and softwares, and

what data you want to visualize, of

course. And then lastly, if you want to

add in actions, you can also do it that

way by just prompting Cloud Code or

Cloud Co-work, but it's important to

tell Cloud or any AI provider to use

Cloud in headless mode, which I think is

also possible for OpenAI or or most

other models, but headless mode in Cloud

basically makes it run Cloud

autonomously in the back without using

the API. So, it will run locally. This

is how we make sure that if you, for

example, want to take an action here by

running an agent or a skill, that you're

not running that through the API, which

of course, again, will cost

significantly more. So, again, you can

just tell Cloud any buttons or actions

you want to include. All you want to

specify is that you want to run Cloud in

headless mode. Now, again, it will

probably take some iterations to get to

a good dashboard, but again, it is very

powerful in my opinion. And then lastly,

I'll show you how to set up a custom

page or app that you can also deploy on

a custom website with password

protection and share with other people,

just like the one I showed you at the

beginning of this video. Now again, the

process of setting this up is a bit more

technical because we're basically

building an a custom app. Uh so first,

I'll give you the overview of how you

would do it if you want to use our skill

that helps you set this up fast, and

then I'll tell you how to do it without

our skill. Now firstly again, because it

lives in a custom interface, we need to

actually set up an MCP just like with

the live artifact out of our second

brain. So you can use the same process I

mentioned in the live artifact to do

this, or again, you can use um the OS

NCP skill in my accelerator

that helps you set this up for you in an

easy way. Now once you have that MCP set

up out of your second brain, you can use

that same plugin that I showed you

before, the Obsidian set set up to get

to an initial live dashboard fast, too.

So you can import that Agentyc OS plugin

in the customize tab, and once you've

added it, you can just use a skill in

out of the plugin, which is the Agentyc

OS setup skill. It'll then ask you if

you want to set this up in Obsidian or

as a custom standalone web dashboard. In

this case, you would select standalone

web dashboard. It's then going to walk

you through the entire process. It will

ask you for your API key or Anthropic

API key, because remember with this

custom dashboard setup, we're using

Claude or any other AI provider we want

with an API of course. You can get your

API key by just going to

platform.claude.com,

log in with your Claude account, and

you'll get an API key. It'll then ask

you some questions like the name you

want to give this dashboard, if you want

to add in any team members, it will ask

you which softwares you want to

integrate. You will of course mention

your second brain connector here, too.

It'll ask you some more questions to try

and personalize your dashboard for you

right away. And after going through some

questions, it will give you a local link

where you can basically test this app

yourself. And if you ask Claude in the

chat to also build you a production app

right away that you can share with other

people, it will also give you that link

right away. You can then open it up and

you'll have your initial dashboard set

up. Of course, connected with your

specific connectors that you indicated

and any specifications you gave it. Now,

it will be in my style of my brand, but

you can easily adapt this by just

telling Claude in that same chat to

adjust the style, maybe throwing your

brand guideline. And of course, this

will just help you to get to an initial

setup. Um the way to actually get this

really personalized and good for you is

by just going into that same chat and

iterating with Claude, asking it to make

changes in the dashboard, adding in

specific actions, etc. Any changes you

want to make, you can just go in this

chat and iterate together with Claude.

And if you want to share it with

someone, uh you can just share that link

and you can also set up a a password

protect by just asking. Now, you'll have

to do this in the code tab again because

it has to deploy things on Railway. And

if you want to do it yourself, again,

first step is setting up an MCP out of

your second brain with the MCP builder

skill, deploying it on Railway, creating

a connector out of it. Then, you need to

work in Claude code and basically build

a custom Next.js or React app and build

together with Claude exactly the app

that you want. Now, when you're building

this app or dashboard, it's important to

know that in this case, you want to

mention to Claude that this has to be

integrated with the Claude SDK because

we need to use APIs because we can't use

the headless mode like with the Obsidian

setup. Of course, you want to mention in

your Claude code conversation all the

MCPs, the skills you want to add. Again,

it is a little bit of a process because

you're base basically building a custom

app, um but I recommend trying it out

because it be can become very powerful.

Again, if this is a bit more technical,

I highly recommend maybe starting first

with the Obsidian setup, uh which is uh

probably a better setup anyway if you're

using this for yourself and already

working with Obsidian because we're not

we don't have to use the APIs, so it's

going to be a lot cheaper, too. And if

you're really new to this, you could

also consider just setting up um that

life artifact, which is probably the

easiest way. If you want access to all

of the resources, skills, plugins, and

the full OS course, uh you can check out

my AI accelerator in the first link in

the description. And if you want to

learn more about the second brain setup

and this agentic OS, you can also check

out the video here above.