The $1M+ Solo AI Agent Business (Full Course)

0:00People are charging $5,000 a month per

0:03customer to build and manage agents for

0:06them. This is a startup idea I wish more

0:09people would do. The customer doesn't

0:11touch tokens or models or any

0:13infrastructure. They just get a digital

0:15employee that knows their business and

0:17it gets better every single week. In

0:20this episode, Nick from Orgo breaks down

0:22exactly how to build this business, the

0:25tools, [music]

0:26the stacks, how to onboard a customer in

0:2830 days, in app and how to actually sell

0:31to busy executives, agencies, and law

0:34firms. We also share the full

0:36implementation playbook, Hermes, Claude

0:38Code, memory layers, skills, all of it.

0:42This type of episode isn't shared

0:44anywhere on the internet. This is the

0:46alpha that people keep for themselves.

0:49I'm giving it to you for free. Enjoy the

0:52episode and I can't wait to see what you

0:53build.

0:56>> [music]

1:02>> I couldn't be more excited to have Nick

1:04from Orgo back on the pod. Nick, by the

1:08end of this episode, what are people

1:10going to get out of it? Craig,

1:12everyone's going to learn not only how

1:15to run a solopreneur

1:17agent business, but every

1:20every gap, everything that they're going

1:21to do wrong from the beginning. I'm

1:23going to save them all the time from

1:24having to learn from my from from from

1:26those mistakes that I made along the way

1:28and

1:29at the end of this video, you're going

1:30to know what what offer to bring to the

1:32market, how to get customers, how to

1:35fulfill, what's the stack for the agents

1:38that you're going to build out, and

1:41yeah, I'm excited to just dive right in.

1:43So, So, Nick, this isn't going to just

1:45be like a pie in the sky, I want a

1:47billion-dollar idea here. This is how

1:51you can take advantage of AI agents to

1:54build a business that maybe does a few

1:55million dollars a year, but not just not

1:58just the idea, right? You're going to

1:59actually share all the tactics from A to

2:03Z, so that by the end of this episode,

2:05someone could

2:07obviously like and comment and

2:08subscribe, but you know,

2:10uh go go and start one of these

2:13businesses, right? Exactly. And like, I

2:16think the big thing is for everyone

2:18who's watching the pod, you're probably

2:20already affluent with AI, and you don't

2:22give yourself enough credit. And the

2:25amazing thing is like 99% of the world

2:28has, you know, there's like many people

2:31are so behind on AI, and you you may not

2:33realize how valuable your skill set is.

2:35Like, oh, if you can set up Claude code,

2:37if you can set up Hermes agent, if you

2:39can set up open claw, that's a very

2:41valuable skill that a lot of businesses

2:43don't have time for, and you can

2:45monetize that. So.

2:47All right, I'm intrigued. Let's go. All

2:49right. So, let me start by I'll share my

2:53screen.

2:54Okay, so let's just dive right in. Let's

2:57go into the offer. So, um when you're

3:00starting a one-person agent business,

3:02you need to have you need to remove all

3:05the friction for your customers. Um they

3:07don't want to think about tokens. They

3:09don't want to think about computer

3:11infrastructure, security, you know,

3:14breaking it when it it, you know, fixing

3:16it when it breaks. They they just want

3:18it to work. And so, the biggest thing is

3:21you need to create abundance in your

3:22offer. And what I have found in in my

3:26own personal uh success with this is

3:29offering unlimited agents, unlimited

3:30usage, unlimited monitoring, support,

3:34security, ongoing changes, etc.

3:36And the key here is you might flinch,

3:39cuz you're like, well, how how how do

3:40how's that even feasibly possible? Well,

3:43the way to do this is to realize

3:45the point it's not that the customer's

3:47going to actually need unlimited agents.

3:49They're they're not going to need

3:50unlimited tokens. But they might they

3:53might think they do and in reality they

3:55might think they need five agents, 10

3:56agents, 100 agents.

3:58When really one, two, maybe three agents

4:01goes such a far away and you can get a

4:04lot of juice for squeeze out of just

4:06properly taking the time to set, you

4:08know, one or two of these up. And that's

4:10where your, you know, that's how you're

4:12going to essentially like control your

4:14costs. You're not spending too much

4:15money on tokens and you're going to

4:18charge 5K a month for this and this is

4:20the the offer that I've been running and

4:22it's been working really well.

4:24And yeah, like customers don't really

4:26need as many agents as they might think

4:28they need.

4:29And you're just going to show them as

4:31quickly as possible

4:33the magic behind it.

4:34So,

4:35this is the offer that I've been running

4:37off the rip. And

4:40I guess I'll I'll read a little bit of I

4:41wrote some of this stuff down. The big

4:43thing here is the point is not that the

4:45customer needs infinite agents. They

4:47don't need infinite tokens or infinite

4:49computers. Most customers they just need

4:51one, maybe two, maybe three. They just

4:54need a seamless experience. Like that is

4:56what you as, you know, as your business

4:59as as the solopreneur agents agency,

5:02you're going to come in and you're just

5:04going to remove all the friction.

5:06And the minute that things start to

5:08break, like the business owners that

5:11you're going to be selling to, they're

5:12going to become so reliant, so dependent

5:15on these agents that if something does

5:17start to break, it is very painful for

5:19them. And so in this video like I want

5:21to make sure that I help you make it

5:24very clear on how to prevent those those

5:26those fall gaps so that when something

5:28breaks you have something in a way to

5:30fix it before they even realize it. And

5:34yeah, if if a customer, if they want

5:37constant improvements, how do you keep

5:38up? How do you fulfill? We're going to

5:40be going through all of that in this

5:42video.

5:43Cool.

5:44Yeah, so I mean my big takeaway from

5:46this is you're selling an AI employee,

5:49you're not selling an AI agent.

5:51People need less agents than they

5:53actually think that they need. And you

5:57want to think about unlimited

6:01You know, you don't want to use the word

6:02tokens basically at all.

6:04Um

6:06And you shouldn't really worry too much

6:07about usage.

6:09Exactly. Exactly. Because for them it's

6:12just it ruins the magic. The minute you

6:14say like oh, like you're going to be

6:16paying for X amount of credits and then

6:19they're always going to be wondering

6:20like oh, how many credits do I have

6:21left? And then you're going to be like

6:22oh, and then it's usage-based

6:23afterwards. It's like the more clarity,

6:25the more simplicity you can create in

6:27the offer that it's just straightforward

6:28and easy, the faster time to yes, the

6:31faster you can get building and the

6:32faster you can just, you know, have a

6:34happy customer. So.

6:36Um

6:37Yeah. That's the offer so far. And so

6:40then the key here is you want to go

6:42vertical. So as always, you want to

6:44clarify you're not a commodity, you're

6:46not just selling

6:48um you know, Claude code. You're not

6:49just selling ChatGPT. You're You're

6:52selling a like a vertically specific,

6:55industry-specific agent. Um you're doing

6:58it fast. It shouldn't take longer than

7:0048 hours to get up and running with the

7:02first agent for your customer. And you

7:05need to talk in terms of time, not not

7:07time saved, but actually outcomes for

7:09the business. So like how much revenue

7:11can you generate for the business or how

7:13much, you know, always always business

7:16outcomes rather than time saved. I feel

7:18like time saved is a little

7:20overused and and people are kind of

7:21immune to that these days.

7:23Um

7:24So that's the offer. It's pretty simple.

7:26And I'll just dive into like what we're

7:29seeing in terms of our own experience of

7:30like running this offer. I believe that

7:33as a one-person business, you can sell

7:36these agents into industries and really

7:38just kind of be not only selling the

7:40agents, but also just creating clarity

7:42around AI. Like I think if you're

7:44watching this pod, you understand AI

7:47pretty well. You probably have a better

7:49understanding than most people, and you

7:51might not give yourself enough credit of

7:52how valuable that is.

7:54Um

7:55and to be the person who can create

7:57clarity around all the noise right now,

7:59that alone is valuable. And and then to

8:01be able to couple that with the tools to

8:03help solve problems in these businesses,

8:05it's like

8:06you you you're going to become so so

8:09irreplaceable for the business that um

8:12yeah, it's really just going to be like

8:14you and the agents are going to be what

8:15drives the value.

8:17Um so I have some industries here.

8:19In red, I have health care and finance

8:22because I don't think that these are

8:25necessarily the best industries to start

8:27off in. They're very high regulatory um

8:29burdens and and red tape. And so um I

8:33actually recommend these other

8:34industries that we're seeing work really

8:36well, which is marketing agencies, law

8:39firms, insurance agencies,

8:41manufacturers, wholesalers, and real

8:44estate agencies.

8:46Um

8:47The reason for these industries that you

8:49might notice is that they're relatively,

8:52you know,

8:53I would say maybe like legacy

8:55industries, not not not necessarily

8:56like, you know, new fast-growing

8:59industries, but they want to be

9:01fast-growing. They want to adopt AI. And

9:03they want they have a lot of pain to be

9:04able to use it as a tool to

9:07essentially just grow their business.

9:09The common pattern with all of them is

9:11they want to be a full-stack AI company.

9:14Meaning they want to be like fully

9:16automated with AI. That's the dream

9:19outcome.

9:20We're not there yet. Um

9:22but you can certainly come in and start

9:26solving the problems from the executive

9:28level, and then it will ripple its way

9:30throughout the rest of the uh the

9:32business.

9:33Um

9:34and I'll dive in on some of the common

9:35patterns, but um how we feeling so far?

9:39Yeah, I think

9:41those are all people businesses. So,

9:43there's a lot of people. When you have a

9:45lot of people,

9:46there's a lot of waste in terms of

9:48efficiency and there's ways to automate

9:50things. That's one. Two is

9:53those uh

9:56a lot of these companies want to be AI

9:58native is another way to say what you're

9:59saying,

10:01but they don't know how. They might have

10:04pieces of their companies that have

10:07become AI native.

10:10They might work with Deloitte. They

10:11might work with, you know,

10:13different, you know, AI transformation

10:15agencies, but, you know,

10:18to assume that these companies are 100%

10:22AI native

10:23is

10:25insane because they're not.

10:28And then the last thing here is

10:32these categories are large, right? Like

10:34law, that's really large. Insurance

10:36agency, that's really large.

10:37Manufacturing, that's large.

10:39Wholesalers, large. The key here is once

10:43you've identified

10:45a category that you want to go after,

10:47then you have to figure out what is the

10:49subcategory or sub niche that I want to

10:51go after. It's too hard to just focus on

10:55wholesalers. But, whole you know, the

10:58way to think about it from a framework

10:59perspective is,

11:01you know, pick a category and then uh

11:05you can you can do like, you know, real

11:07estate agencies in Florida. So, that's

11:10like, you know, a geography is one way

11:12to do it. Or you can pick a

11:16specific type of real estate, uh

11:20you know, professional.

11:21So, it could be commercial real estate,

11:24you know, agencies in Florida. So,

11:27there's there's different ways to think

11:29about how you can

11:30niche down. And that's going to be

11:33really key here because if you want to

11:35create an irresistible offer,

11:38uh you know,

11:40a big way to get the attention of

11:42someone is to be like, "Oh my god, this

11:45person is really speaking to me."

11:48Exactly. And honestly, like even a

11:51little bit of a like some maybe some

11:52contrarian advice from my end is like

11:55you don't have I

11:56I have a feeling as though you don't

11:58have to start super niche from the

12:00beginning. In fact, you can always niche

12:03down after trying a marketing agency,

12:05trying a law firm, trying all these

12:07different industries, seeing what works

12:09well for you, where the market pulls

12:10you, and then going super vertical. But,

12:14um I really love like the concept of

12:17like it's a design thinking principle of

12:18diverge and then converge. So, like you

12:21know, try a try many different things as

12:23long as it's not, you know, too for too

12:25long cuz you don't want to get into this

12:27constant um cycle of trying something

12:30new and, you know, you never get to

12:31focus. But, once you find the thing that

12:34clicks for you, whether it's you're able

12:36to resonate with the audience really

12:37well or you're just getting pulled into

12:39that market more, like yeah, go super

12:41niche, go sub niche, and use that as

12:44your wedge to kind of like infiltrate

12:46the rest of the market. Um

12:48yeah, I think that's spot-on.

12:50So, and then as far as the common things

12:53that we're seeing, right? So, with all

12:56these industries, what you'll find is

12:58the people you're going to jump on calls

13:00with, the people that are going to

13:02likely be the decision makers and the

13:05ones purchasing your service or your

13:07productized service, these are the

13:08executives, these are the decision

13:10makers. And when you abstract on all

13:13these industries,

13:14the decision maker at the end of the day

13:16has very similar problems, no matter

13:18what the industry is. They have too many

13:20emails, too many meetings, too many

13:21follow-ups, too many open loops. They

13:24have context over so many different

13:26projects and places and people to keep

13:28track of.

13:29And so just out of the gate, if you can

13:31anticipate this, you can have something

13:34that you put together that, you know,

13:36maybe from a template perspective solves

13:38a lot of these issues. And then you can

13:40cater more specifically into that niche,

13:42into that vertical for that industry.

13:45Uh, if it's a if it's a a law firm and

13:47you have a partner who wants to buy your

13:50services,

13:51um, you can have all of these things out

13:53of the blocks for your agents that you

13:55set up, which I'll show you how to do

13:57also in this video. And

14:00and and then you could also cater it for

14:02that particular industry. So, oh, yes,

14:05we have a agent that does, you know, um,

14:08you know, following uh, following up

14:10with people, projects, etc., but it also

14:12manages your cases, it does demand, um,

14:17demand letters for your law firm, it

14:19does, um, all the different things and

14:21skills that you would need for for, you

14:23know, um, maybe a matrimonial law firm,

14:26for instance. So,

14:29that's the uh, abstraction layer on no

14:32matter what, you're going to be solving

14:34a lot of executive problems, and then

14:36the key is to layer in uh, vertical

14:38specific solutions as well.

14:40Um,

14:42okay, so there's that. That's the

14:43market. So, we talked about the offer,

14:45we talked about the market.

14:47And I have some side things as well

14:48about like

14:50how to get customers. At the end of the

14:51day,

14:53um,

14:54I think everyone should make content.

14:56>> [laughter]

14:58>> I think if you if you if you can jump on

15:00a call, this is just a little tidbit, if

15:03you can jump on a call with somebody and

15:04they know who you are and what you sell

15:06without you having to tell them and

15:08they're warm to begin with,

15:10that's the ideal position to be. You

15:12never want to be in a position where

15:14you're, you know, having a cold call.

15:17You never want to sell to a cold

15:18audience. And, um, you know, in the

15:20beginning, you might have to, so, you

15:22know, starting for free even is

15:23sometimes worth it just to get case

15:25studies and get referrals. But um

15:28content is like overpowered in 2026. So,

15:31I do recommend that. I mean

15:34that's how we met, you know, it's like

15:36it's like midnight, can't fall asleep.

15:39I'm like doom scrolling Instagram, I see

15:41Nick's face pop up showing me, you know,

15:45how to use open claw and I was like this

15:47is a guy who has some sauce and I need

15:50to have him on the podcast.

15:52So

15:54the other thing about creating content

15:56is not only is it helpful in terms of

15:58getting your face in front of customers

16:02or or your offer in front of customers,

16:03but it also helps you you know, get

16:05known, get on podcasts, hire the right

16:09people. So, there's there's a lot of

16:11advantages in an AI world when you can

16:13use AI to automate a lot of the research

16:17and a lot of the um just help you know,

16:20the editing and things like that.

16:23Just do it. Like I hate to say it, just

16:25just do it. You got to just do it. It's

16:28just like it's the most leveraged thing

16:30you can do. It's like content and then

16:33if you think of other like leverage

16:34things, it's like okay, AI or you could

16:36also have leverage with talent and

16:38software, but um

16:41yeah, it's it's it's incredible and I

16:43think like the trend of 2026 is

16:46content is king

16:48and

16:50and I'll tell you a little bit of a

16:51tidbit as we go into this next segment,

16:53but I don't know about you, Greg. I've

16:56been going on walks and what I'll do is

17:00I'll go on a walk and I'll I'll send off

17:02a long

17:03um horizon task to my agent via

17:05Telegram. I have my own, you know,

17:07Hermes agent is what I use these days.

17:09And

17:11it's I'm just like I'm just in awe with

17:14what the world we live in today. Like

17:17how amazing I can go on a walk and

17:19there's work being done on for for

17:21business and on customers and you know,

17:23for their agents by my agents. And I'm

17:26just like, if you extrapolate that over

17:28the next 6 months, 12 months, like

17:32the the most leveraged thing you could

17:33do is post a piece of content that

17:36reaches a lot of people and then have

17:39this robot that helps you fulfill for

17:42the thing that you're providing as you

17:44go on a walk or you right before you go

17:45to bed or when you wake up. It's just

17:47It's amazing. It's an incredible world

17:49we live in. So, um yeah, let's dive into

17:52this stack, shall we? How do we build

17:54these things?

17:56Okay, so as far as the tools that you

17:59might need to fulfill for your service

18:03of, you know, providing agents for

18:04businesses, um first and foremost, I use

18:08Granola. I love Granola. I use it for

18:10every meeting. They have an MCP, you

18:13know, you can give it to your agent and

18:14then it just has context over

18:16everything. And what I do is these

18:18meeting notes from Granola, they

18:20automatically get synced into our

18:22requests on Trello.

18:24Um and so Trello is the customer-facing

18:28um like essentially project management

18:30um Kanban board that I use. And so, you

18:33know, there's a backlog list, there's a

18:35to-do list, there's a doing list,

18:38there's a a done list, and the customer

18:41can just simply drag and drop what they

18:44want into the to-do list for, oh, I want

18:47my agent to be connected to

18:49um my calendar. I want it to have access

18:52to this other platform. I want it to

18:54create content for me. And they can just

18:56add these requests at um one at a time.

18:59And the key here is

19:01these agents can at this point do so

19:04many different things. They could do so

19:05many things that um you almost need to

19:09create, you know, prevent scope creep

19:12by, you know, limiting one to two

19:13requests in under 48 hours

19:16because there's a lot and you could do a

19:18lot, but you just need to be careful

19:20that you don't, you know, end up

19:21drowning in a fulfillment nightmare.

19:23Um, so that's why Trello's helpful in

19:26terms of scoping.

19:28Loom is awesome. Your customers are

19:30going to want you to send them updates,

19:33you know, send an update at 2:00 a.m.,

19:34send an update at, you know, different

19:36times of the day of you implementing new

19:39things for the agent and whether you

19:40improve the memory or you improve the

19:42the Obsidian Vault that it's it's

19:44operating off of off of.

19:46Um, Loom is awesome. And then I just use

19:48like Calendly link. Like I have a

19:49horrible funnel.

19:51But you can do you can do a lot of

19:54um

19:54you can get a lot of bookings. Just

19:56Calendly link, personal website, drive

19:58traffic there, create content. Um, these

20:00are like pretty much the customer-facing

20:02tools. And then I have

20:04I don't know about you, Greg. Do you use

20:05Do you use Superhuman?

20:07The email tool? Yeah. I I don't, but

20:11people tell me I should. Oh my god. I If

20:13you If you have a lot of emails, you're

20:15going to have a lot of emails with

20:16customers, oh man, it's Superhuman is

20:19amazing. It has a bunch of shortcuts. I

20:21love keyboard shortcuts. And you just

20:24fly through emails.

20:26Um

20:26and it's not like it's AI generated.

20:28Like it makes you write the email and

20:29you have AI help you, but it's just a

20:32very focused uh focused platform. And

20:34then lastly, Asana. I use Asana for

20:36internal facing. So not customer facing,

20:38you know, if I want to keep track of

20:40some specifics or uh details of of what

20:43needs to be done. Um,

20:44yeah, that's the that's the software

20:46stack.

20:47Um

20:49Okay, let's dive into the the agent side

20:51of things now. So

20:53for building agents, the irony here is

20:55if you don't know how to build an agent,

20:57please don't worry.

20:58I got you.

21:00We're going to use agents to build

21:02agents.

21:03Um,

21:04and so Claude code, they have a new

21:07desktop app. It's awesome.

21:09Open AI's Codex, they have a new uh

21:11desktop app. It's awesome.

21:14And you can actually use these to build

21:16the agents for your customer.

21:18Um, and as far as what agents to use,

21:22you have a couple options. You're not

21:24going to sell Claude code to your

21:26customer. You could, or Codex. I mean,

21:28you could.

21:29But I highly recommend using Hermes

21:32these days. I find it to be the most

21:34reliable. Um, it allows you to pick any

21:37model. The reasoning here is

21:39tomorrow there's going to be a new model

21:41that comes out. It's going to be

21:42infinitely cheaper and it's going to be

21:44Opus 4.7 level intelligence and it's

21:46like, you just want to have the

21:48flexibility to quickly switch whatever

21:50the agent that you're running it uh

21:53whatever model it's running. Be able to

21:54switch that quickly and you don't want

21:56to be married to a platform, married to

21:58a tool, married to a

22:00an infrastructure. So,

22:02um Hermes, I really like. Have you

22:05played around with Hermes at all? I I

22:06think I saw some videos. Yeah, yeah, I

22:09have. Um

22:11I haven't, you know, quite made the

22:13shift yet.

22:15Uh but uh I've done an episode on it

22:18with my friend Imran. So, go check it

22:20out if people are interested in learning

22:21about how to set up Hermes.

22:23We called it Hermes cuz we're fancy like

22:26that in the episode and then we got

22:28you know, the team at or Hermes, you

22:31know, quickly corrected us.

22:33They did? Oh, wow. I I like Hermes.

22:36Yeah, Hermes is a little more fancy. If

22:37you if you sell Hermes agents, you can

22:39charge 10K a month. Exactly. So, uh open

22:44clause commoditized it already, you

22:45know. It's 5K a month. It's okay. So,

22:49you pick your harness here. So,

22:51this is, you know, the agent that you'll

22:52sell and you need a place for that agent

22:55to live. You can use something like

22:56Hostinger, you can use Origo, you can

22:58use

22:59whatever. Um,

23:01I obviously I'm biased um but Orgo is

23:06really nice because in one workspace you

23:09can have all your agents, have your

23:11agent managing their agents and I'll

23:12dive into all of this

23:14and getting set up. And then lastly

23:17you need the tools for the agents. So

23:19some things out of the box that I

23:20install for every agent no matter what

23:23outside of just giving them a computer

23:25and the ability to use it is Composio.

23:28Have you heard of this company Composio?

23:31I have but can you give a one-liner for

23:33folks who haven't heard of it heard of

23:35them?

23:36This company allows you to this

23:39connector they they allow you to have

23:41one connector one MCP essentially that

23:45connects to thousands of other apps

23:47whether it's Gmail, Slack, Notion what

23:49have you. And with one connection you

23:52can manage you can have access to all

23:55the tools that you would need to send an

23:57email via Gmail or push something via

23:59GitHub or pull a message via Slack.

24:02It's incredible and it handles the tool

24:05the tool calling and the authentication

24:07which is huge because security is like

24:10the biggest challenge with setting up

24:12these agents like

24:14by far the biggest time sink is getting

24:17authentication set up for the customer

24:19cuz you have to uh what's your username

24:21and password for this and then

24:22if you email it it's like not secure so

24:25then you use something like Composio

24:26done.

24:27So it handles that and then it handles

24:29security in that sense as well

24:31everything's managed through their

24:32platform and then it manages

24:34handles the tool call. So if you have

24:36Composio set up with all the connectors

24:38you can just take that one connector

24:40take it to any agent and it has all the

24:41same connectors. So I really like this

24:44company I don't have any affiliation but

24:46I love their product. Um

24:49really great.

24:50Next up is agent mail.

24:52This one is I give every agent uh, an

24:55email.

24:56It adds a nice personal touch. So,

24:59um, you know, let's say you're you're

25:02you're an executive, I give you an

25:04agent, you name it Mia,

25:07and Mia needs Mia needs her own email.

25:11Agent mail allows you to give Mia an

25:13email so that she can send or receive

25:14emails, and that's really fun cuz it's

25:16it turns into like truly like a personal

25:19assistant.

25:21And then lastly, Obsidian.

25:23Um, you have a video on Obsidian it did

25:25really well. Obsidian super important

25:28because at the end of the day these

25:30agents need context, and the more

25:33context you can provide in a nicely

25:35wiki-styled structured format and

25:38markdown files for the agent,

25:40um, it will really just thrive in terms

25:42of understanding projects, people,

25:45things that you're doing, so on and so

25:47forth. So, um, this is the stack, and as

25:52far as models, like as final touch

25:53around models, um, today, by far the

25:57best model to use for something like a

25:59Hermes agent or an open claw, is GPT

26:025.5.

26:03Um,

26:04it's so efficient with the tool calls,

26:07it doesn't eat through tokens like Opus

26:094.7 from Anthropic does, um, and and

26:13OpenAI is very generous around letting

26:16you use your paid plan uh, with with

26:19with any model like with any harness

26:21like like Hermes or or open claw, and

26:24then, um, and you just get a lot of

26:26usage out of it. So, I recommend 5.5. If

26:29you want to use open-source models that

26:31are a little more affordable for

26:32lightweight tasks, uh, GLM 5.1 from Z AI

26:37is, in my experience, like the best

26:39open-source model to be using. Kimmy

26:41comes in on a on a close second,

26:44uh, and these are both more affordable.

26:46Uh, and then

26:48Opus 4.7 finally, if you have some long

26:51horizon coding task, um, Opus 4.7 is

26:54really great for that and

26:57you can actually have your agent

26:59connect to Claude code and be able to do

27:02these long coding tasks in Claude code

27:04and then bring that back to the agent.

27:05So,

27:07tidbit there.

27:09I want I don't know if you can do this

27:11real quick, but could you give a

27:13one-liner on because people are going to

27:16sorry, let me take a step back. People

27:17are going to look at this list and

27:18they're going to be like, "Oh my god, I

27:19don't know if I should I should use

27:20Codex or if I should use Claude code or

27:22if I should use Open Claude or if I

27:23should use

27:24Hermes, should I use Hosting or should I

27:27use Orgo or should I use this or should

27:28I use Can you go and just quickly, you

27:31know, what's NYX stack and like with a

27:35one-liner of why you use that tool over

27:37the other tool?" Yeah.

27:39Codex because it's more generous and

27:43it's simplest and they have the best

27:44desktop app. Hermes because it doesn't

27:47break and it's self-evolving. Open

27:50Claude is not as self-evolving.

27:52Orgo because we give your agent a

27:55computer so it can live in the computer,

27:57it can operate the computer.

27:59We're not just a headless VPS server in

28:02the cloud

28:03and I'll dive in on that.

28:05Composio, you need this. Everyone needs

28:07this.

28:08Agent mail, everyone needs this.

28:10Obsidian, everyone needs this

28:13and then

28:13>> Wow, that's a hot take by the way.

28:15Obsidian, everyone needs this. Oh, yeah.

28:19>> Yeah,

28:20good just I mean, explain why you should

28:23use Obsidian over say Notion.

28:26So, here's my Obsidian vault

28:29and

28:31Greg, I I've been I've been building

28:33this vault since November of 2025. Agent

28:37I mean,

28:38what do you How do you How do you say?

28:40Um

28:41when something's outdated

28:43like super old. 2025 is forever ago, you

28:47know? Um so, I've been building this

28:50vault since 2025 November, before Open

28:52Law, before Hermes. And it has

28:54everything about people, projects,

28:58everything.

28:59And I'm so crazy,

29:01I have

29:03uh a limitless microphone.

29:05Even that, daily transcripts get pulled

29:08from that into here.

29:09Um

29:11this is genuinely a second brain. Like

29:13Like people say Obsidian that's a second

29:14brain and then

29:16okay, they they show it. Okay, that's

29:17kind of cool. They use it for some

29:18research. No, no, no. This is a second

29:20brain. And when you have something like

29:22this, it is quite literally

29:25you get to experience what personal

29:28AGI might feel like uh in the next 3 to

29:316 months

29:33uh from now. I'm sure everyone will

29:34experience it, but I feel like I'm

29:35getting to experience it sooner because

29:37I just have such well-organized

29:40uh markdown files.

29:41Um

29:42It's incredible. It's incredible. So,

29:46it just gives your agents context on

29:50what it needs to know given any given

29:52tasks. And it feels like it just never

29:55forgets and it understands you.

29:58Um

29:59And I think that's at the end of the day

30:00like we just want an agent that

30:01understands us and

30:03helps us with our business and and just

30:05has perfect context over everything we

30:07do. So,

30:09Enough said.

30:10Yeah. And then 5.5 is the best. I would

30:13just use 5.5 to make it easy. Yeah, GPT

30:155.5.

30:17So,

30:18as far as that's the stack,

30:20um

30:21now, in Orgo, we give the agent a a

30:25computer to live in. Greg, let me invite

30:27you to this.

30:28Um

30:29All right, so I just invited you, Greg,

30:31into this workspace in Orgo. And we're

30:33just going to quickly spin up a computer

30:36here.

30:36And I'm going to spin up a computer. I'm

30:38going to say

30:39um

30:40Greg's computer.

30:42I'm going to launch it.

30:44And it launches pretty fast with really

30:46fast desktops. Um,

30:48and now that we're in this workspace

30:51here, in this computer,

30:52I can now install the agent inside of

30:55it, actually. So, the agent can live

30:56inside of here. So, if it's Open Claw,

30:58if it's Hermes agent in this case, it

31:00will live inside of this environment.

31:03And, um,

31:05the key here is, regarding to getting

31:07set up,

31:08we have an Orgo MCP that is what I

31:13use for for setting up agents. And that

31:16little story I told earlier about going

31:17on a walk and be able to get work done

31:20on a walk, it's because my agent

31:23is using a Orgo MCP to connect to my

31:26customers' agents that live on Orgo.

31:29And so, what ends up happening is Orgo

31:31is like this workspace

31:33where my agent and other agents and

31:36myself can all collaborate on these

31:39computers where these agents live and

31:41get them set up and configure that way.

31:43Um,

31:44so,

31:46here I have a, um,

31:48I don't know if can you see my Telegram

31:50chat?

31:51Yep.

31:53I had my agent last night, actually. I

31:54kicked off a task. I told it to, uh, go

31:57ahead and build out some CLI and skills

31:59for, uh, for for Orgo. Um, but I'll

32:03start a new chat here and I can actually

32:05just tell the agent,

32:07um,

32:07I'll grab the computer ID from Orgo. So,

32:09let's grab this computer ID.

32:12And

32:13I can give this computer ID to the

32:15agent.

32:17And I'll say, "Set this computer

32:21on Orgo up.

32:23Um,

32:25computer ID quoted here.

32:28Let's install Hermes

32:30agent into the VM."

32:34So,

32:36the reason why I tell everyone not to

32:39get stressed out or scared about setting

32:41up these agencies,

32:43you really just need another agent to

32:45set it up. In my case, I'm using another

32:46Hermes agent to set up

32:48a Hermes agent.

32:50In another case, you could

32:52I'll spin up another computer here. In

32:54another case, you can literally install

32:57something like Claude Code into a VM on

33:00Orgo.

33:02And you can actually just run Claude

33:04Code from the terminal here and tell

33:06Claude Code in natural language, "Hey,

33:09let's set up Hermes agent." Um so, just

33:12real quick, you know, you go Claude Code

33:14install command

33:15for Linux.

33:17You find that real quick.

33:19And you just run this in the terminal

33:21here and you would literally install

33:23Claude Code, run it from here, and have

33:26it install Hermes into this VM.

33:29Um

33:31So, the answer to all of our problems,

33:33Greg, is that

33:35more agents is the answer. Uh if you're

33:38confused on how to set something up,

33:40have your agent do it.

33:42Um

33:44And yeah, so I'm just going to install

33:46Claude Code here.

33:48It's going to get that going.

33:50And we'll be off to the races.

33:53This will be a dumb question,

33:55but why are we

33:58doing virtual computers versus doing

34:00local computers? You know, buying Mac

34:03minis and and doing that whole thing.

34:06It's actually a very, very good

34:08question. And the reason is we want the

34:10ability to

34:13work on our customers' computers from

34:15where we're at. And if you are using a

34:18Mac mini,

34:20k-

34:21I can't even imagine the nightmare of

34:22having to go in person and, you know,

34:24debug something that's like at a

34:26hardware level or

34:28um something on the Mac mini breaks or

34:29an update or or what have you.

34:32Orgo gives you cloud computers to be

34:34able to manage these agents. And um with

34:37that, you can do so many more amazing

34:39things both from a perspective of just

34:41scaling your business, being able to

34:43access all these agents on one platform,

34:46via one connector, and have your agent

34:48connect to all of them.

34:50Um that's really like the biggest thing.

34:52It's just from a fulfillment

34:53perspective. Um

34:56it is just like the easiest way. And

34:57then also just a security perspective.

35:00These are isolated cloud computers, and

35:03you can delete them in under a second.

35:05You can create a new one.

35:06And with that, there's a lot more

35:09sandbox environments that you could just

35:11protect you and your customers from

35:14uh like a blast radius that might

35:16otherwise be

35:17um more dangerous on a personal Mac

35:19mini. So. And so say I have 100

35:23customers. Am I creating like how am I

35:26structuring that? Am I creating like

35:28separate like projects with these

35:31computers in it? Like what is from a

35:33best practices perspective in terms of

35:35security and just, you know, good UX,

35:39how should people think about setting

35:41that up? Yeah, so in this case, it would

35:44be exactly like you said. Like I if you

35:46were if you were a customer,

35:49um I would just make a workspace for

35:51your business, and I would say,

35:53um you know, this is let's do a like

35:56idea browser, right?

35:59We would create this workspace, and each

36:02each of your agents would live in this

36:04workspace, and then I'd have other

36:06workspaces for other customers, and I'd

36:09be able to manage all of that, you know,

36:10on on Orgo. Um

36:14yeah, one platform.

36:17Cool. Yeah, I think what's also cool

36:19about this, just just how visual it is,

36:21like

36:22showing this to a customer and being

36:24like, I know you think,

36:27you know, it's not secure. You might

36:29think it's not secure or you might

36:30think, you know, but this is

36:32you know, a visual sandbox environment,

36:34right? Like it's just it feels like the

36:36cell. Like you just like

36:38You know, you talked about Loom before

36:40but like showing looms of this, I think

36:42it's just going to light people up.

36:44Yeah, exactly. And also like you can

36:46also

36:48out of the box on Orgo like we have this

36:50playground mode here. So like this is

36:52our this is just all of the latest

36:54models from, you know, Anthropic and and

36:56Kimmi and and Chat GPT. And so as far as

37:00a demo, when you tell a customer like,

37:02"Oh yeah, like we can we can have an

37:04agent like operate a computer and do do

37:06things for you." And essentially you're

37:08describing Hermes agent or you're

37:10describing Open Claw. Even then they

37:12they might have a hard time like

37:14imagining like what what what does that

37:16look like? What does that feel like? And

37:17so when you just give it a computer,

37:19you're able to just give it life and you

37:21can tell them like you can say like,

37:22"Hey, look up what is Idea Browser

37:26and search it on Google."

37:29And

37:30this actually becomes like a really good

37:32demo. Like

37:34for for you and your customer like to be

37:36able to show, "Look, oh, the agent is

37:38controlling a computer

37:41and it's doing research and it's doing

37:43real work.

37:44And you can just quickly show a demo in

37:46Orgo. It's like super cool.

37:48Um

37:49This is cool. Yeah, like even me as like

37:52a co-founder of Idea Browser, I'm like

37:54looking at this and I'm like, "Yes."

37:56Yeah.

37:58Yeah, it's awesome. And and oh, we have

37:59some I'm I'm I'm making Jordan some

38:02agents. I don't know if he told you. I'm

38:03making him some Idea Browser agents and

38:06he's uh he's using them. He's stress

38:07testing them. Um

38:10but everyone needs agents. So it's um

38:14yeah, it's really cool. And then as far

38:16as the Telegram set up here like

38:18you can see

38:19my agent is literally using Orgo MCP to

38:22get this Greg's computer set up right

38:24now. And it's installing Hermes agent.

38:26Um you know, part of this yeah, these

38:29some of these things take time. It's

38:31running a long process. It might take

38:345 10 minutes. So, I'm not going to bore

38:36you by sharing that, but um the concept

38:38of use agents to set up other agents,

38:41it's very real. Um and I can also

38:44dive into practices around that and how

38:46to make sure that your agent knows how

38:49to set up other agents. Um

38:52yeah.

38:53Does that sound good? Let's do that.

38:55That sounds great.

38:57So,

38:58to have your agent have context of how

39:00to set up other agents, it actually

39:02needs I wish I would have added them

39:04here. Um

39:06a few MCPs go really far away. One of

39:08them

39:09is the Perplexity MCP.

39:12Um

39:13with Perplexity, you can give your

39:15Claude code or Codex

39:18uh real real up-to-date knowledge on

39:20things like Hermes. And so, the key here

39:23is like you always just want to have

39:25any sort of setup process,

39:27initialization process grounded in real

39:29context of what is the docs for setting

39:33up Hermes agents today and what how do I

39:35connect my Hermes agent to iMessage? If

39:38you have something like Perplexity, you

39:40give your agent the ability to see how

39:42to do that and be able to set it up

39:44perfectly. Um Xa AI is another

39:48great MCP tool for like real-time web

39:50search.

39:52Another big one actually is Context

39:53Seven.

39:55This one is awesome for getting

39:57up-to-date docs from like GitHub from

40:00like Hermes agents GitHub, so they they

40:02can see specifically the docs of how to

40:05get set up.

40:06You just need some sort of context layer

40:08to let to loop in the best practices and

40:10up-to-date docs

40:12um for setting up these agents. And kind

40:14of like one final recommendation would

40:16be the XMCP. So, Twitter released their

40:19own MCP and I find so many amazing

40:23setups on Twitter for open claw and

40:24Hermes agents that there's many times I

40:26just want to use

40:28that context for setting up an agent for

40:31a given task and you could actually use

40:33this, give this to your Claude code or

40:34give this to your Codex and have it

40:39use this context to help help you set

40:40things up

40:42or you could use all of them, too. So,

40:45I mean, is there any downside to using

40:46all of them?

40:48Um no, I I use all of them.

40:51And so, like maybe even in here when you

40:53look at my Telegram,

40:55you might notice oh, I guess it's it's I

40:57didn't ask it to to pull in up-to-date

40:59context. Mine has skills already built

41:02in place to be able to like set these

41:03things up just cuz I do it so much, but

41:07in general, like the more the merrier.

41:09Like context is key and

41:13I like to like have sub agents spawn.

41:16I'll tell Codex like, "Hey, or Claude

41:18code, hey, spawn five sub agents. One

41:21sub agent for Perplexity, one for Exo,

41:23one for Context.ai, one for Firecrawl,

41:25one for

41:27um

41:27XMCP

41:29cuz I like to pull from different

41:31resources and then those all come back

41:33to the main agent and we get the best

41:35practices. So, that's how I do it.

41:40Cool.

41:42Let's see. Okay, so this here

41:44is the

41:47I should have ran this in the terminal

41:48below, but I'll go ahead. I'll I'll just

41:50run it in this terminal.

41:52And what we're going to do here is

41:55do that, run Claude.

41:57Or you could just like spin up multiple

41:59terminals. [laughter]

42:01Yeah.

42:03You can Oh, wait, let me see.

42:07Okay, I think I just copy this here. So,

42:10Okay.

42:13Oh, command not found.

42:16Okay, I'll I'll I'll I'll debug this

42:17later.

42:19Um

42:19This is also why This is mainly why I

42:22use the Orgo MCP. I'm just like I let my

42:25agent do all the work. Um actually, you

42:27can also come here into the playground

42:29and say

42:30install Claude code into this computer.

42:33Um and just have our agent do it cuz I I

42:36I'm

42:37I don't want to debug what's going on in

42:38the terminal right now. So, then just

42:40have this agent do it um and do it that

42:43way.

42:44But yeah, once you have it set up from

42:46here, like

42:47I can now ask my Telegram agent. Um so,

42:52I'll I'll stop this here.

42:54You can just imagine, you know, when

42:56that's done, your agent's set up. And

42:59I'll start a new chat. And I'll ask

43:01like, "How many Orgo VMs

43:04do I have in my workspaces?"

43:08And my Orgo claw

43:10uh is able to actually manage all of my

43:15customers' agents from, you know, just

43:19this one agent. And it can upgrade, fix

43:21things on the fly, you know,

43:24and

43:26all from one spot. Anywhere I'm at. Um

43:29if I if I get an email from a customer

43:31that uh something broke,

43:34we can just send off an agent uh send

43:36off a message to Orgo claw and have it

43:38go fix it. Um

43:42Boom. You can see here, 27 Orgo VMs

43:44across your workspaces, all 27 shows

43:46running.

43:49And then it dives in onto all the

43:51different customers and all their

43:53agents.

43:54Um

43:55So, last point that I want to make

43:57around getting these things set up is

44:00the

44:01uh watchdogs. So,

44:04the gateways are what make these agents

44:07connect to a platform like Telegram or a

44:10platform like

44:12WhatsApp and sometimes these gateways

44:14crash. Open Claw

44:16has a lot of gateway issues in my

44:18experience. Hermes is a lot better.

44:20Um and so a key here is you want to make

44:23sure that you set up a um watchdog. You

44:27could literally just tell your agent set

44:29up a watchdog for whenever a gateway

44:31crashes that it auto restores it. Um

44:35that's super important just from, you

44:38know, reliability perspective. A second

44:40thing is you want to make sure that you

44:43have some layer of observability or

44:46alerts. So, I have agents email me

44:50if I set up your agent, your Mia agent

44:53and Mia has an email

44:56Mia's Mia emails me from her email when

45:01her cron job breaks or her skill failed

45:04or something happened and I'm alerted

45:07about it and I can go in and then debug

45:10it and fix it

45:12which is super valuable cuz once again,

45:13for your customer, you don't want them

45:15to have to worry about like

45:17doing all this themselves. So,

45:20um

45:21make it as simple and easy as possible

45:23handle everything tip to tail

45:25and

45:27yeah, I think

45:28I guess like the big takeaway here is

45:32it is hard to set up Cloud Code even.

45:34Like people are like Cloud Code is is

45:36going to kill Open Claw or Cloud Code's

45:38going to kill Hermes agent. And in a

45:40general sense, it's getting better at

45:42doing a lot of these general things. But

45:44to be able to go in and create a

45:46specific agent for a specific industry

45:48and person and have it tailored to their

45:50workflow it's like

45:53you're underestimating how much value

45:55that is and you can really create a

45:57lucrative business by yourself. Uh Uh,

46:00and your agent building other agents for

46:02other businesses. Um,

46:04yeah, and I think it's an amazing time

46:06to be a a solopreneur for this.

46:09You can and you will. So, Nick, thank

46:12you for

46:14sharing the playbook for how to build a

46:16one-person

46:18agent-led business,

46:21uh, sharing how to actually do it, um,

46:23in such a clear way.

46:25I love chatting with you because you're

46:29you give the sauce, but you also explain

46:30it super clearly.

46:32Nick is a criminally under-followed

46:34account on on social media.

46:36Uh, he you know, he's getting some some

46:39followers, but I think he can be I think

46:41he needs to be bigger, so I'll include

46:42links for where to find Nick

46:44uh, in the show notes in the description

46:47and uh,

46:49Nick, I'll see you in a few weeks in San

46:50Francisco and let's have a

46:52let's have some coffee and have a good

46:53time.

46:54Thank you, Greg. Always a pleasure.

46:56Thank you so much and um,

46:59I I hope to yeah, I'm going to see you

47:01soon. We're going to get some coffee.

47:02We're going to we're going to do some

47:03sipping time.

47:05>> [laughter]

47:05>> We're going to do some IRL sipping time,

47:07which is my favorite. It's there's

47:10there's nothing like it, you know, like

47:12I actually have been trying to

47:14cut down on my like Zoom meetings and

47:17stuff like that. It's just there's not

47:18There it is. There's nothing like

47:20being in person, sharing ideas,

47:23sipping,

47:24and

47:26uh,

47:27and figuring out what

47:29what we can be building in in in a time

47:32like this because there's so much and

47:34sometimes the hardest part is figuring

47:35out the right idea, the right time, the

47:38right playbook, the right steps, the

47:40right order, and uh, this has been

47:42helpful, Nick, and and definitely got my

47:44great creative juices flowing, so I'm

47:46sure others are very thankful as well.

47:48So, thank you, Nick, and I will see you

47:51next time. Thank you, Greg. Talk soon.