The argument in one line.
Solo founders can build million-dollar businesses by selling AI agents at $5,000/month to legacy industries, handling all technical setup and ongoing management so customers never touch infrastructure.
Read if. Skip if.
READ IF YOU ARE…
- A solo builder or freelancer who can already set up Hermes Agent or Claude Code and wants to monetize that skill by selling managed AI agent services to legacy businesses at $5K/month.
- Someone evaluating which industry verticals to target for an AI agency and wants a specific playbook covering offer design, client onboarding, and stack selection — not theory.
- A consultant or agency owner looking to productize AI delivery with a repeatable unlimited-agents offer that manages token costs while keeping the client experience hands-off.
- An early-stage founder who wants to watch a live Hermes/Orgo demo that shows how an agent-builds-agent workflow actually looks in a real client delivery context.
SKIP IF…
- You have no experience with Claude Code, Hermes, or agent infrastructure — the playbook assumes you can already set these tools up and focuses on business model, not setup.
- You are targeting consumer or B2C markets; the entire framework is built around selling to executives, agencies, and law firms with budget for $5K/month retainers.
The full version, fast.
A solo operator can build a multi-million-dollar agent business by selling a fully managed AI employee to legacy industries at around $5,000 per month, removing every mention of tokens, usage caps, and infrastructure from the offer. The mechanism is a productized service: pick a vertical like law firms, agencies, insurance, manufacturing, or real estate, promise unlimited agents and ongoing changes (most clients only ever need one or two), and fulfill with a stack of Hermes as the harness, GPT-5.5 as the default model, Orgo cloud computers per client, Composio for tool authentication, Agent Mail for identity, and an Obsidian vault for context. Use agents to build and monitor agents, scope requests through Trello, and add watchdogs plus email alerts so failures self-heal before the client notices.
Members feature
Chat with this breakdown.
Modern Creator members can chat with any breakdown — ask for the hook, quote a framework, find the exact transcript moment. Unlocks at T2: refer 3 friends + add your own API key.
Create a free account →Who's talking.
00:00hostGreg Isenberg
01:00guestNick Vasilescu
Where the time goes.
00:00 – 02:54
01 · Intro
Greg opens with the $5K/month hook VO, introduces Nick from Orgo.
02:54 – 06:38
02 · The Offer
Unlimited agents, unlimited usage, unlimited support at ~$5K/month. Customers think they need 10-100 agents; reality is 1-3. Never say tokens.
06:38 – 07:26
03 · Sell an AI Employee
Greg crystallizes the framing: selling an AI employee, not an agent. Magic dies when customers hear tokens.
07:26 – 14:54
04 · The Market
Target legacy industries: marketing agencies, law, insurance, manufacturing, wholesale, real estate. Avoid healthcare/finance. Diverge then converge to a sub-niche.
14:54 – 17:51
05 · Getting Customers
Content is the warm-pipeline engine. Start for free to build case studies. Nick got on this podcast because Greg saw him on Instagram.
17:51 – 20:49
06 · Customer-Facing Stack
Granola to Trello to Loom to Superhuman to Asana.
20:49 – 28:14
07 · Agent Build Stack
Codex/Claude Code to build. Hermes as the agent harness. Orgo for cloud VMs. Composio for auth. AgentMail. Obsidian as second brain.
28:14 – 30:22
08 · Obsidian as Second Brain
Nick vault (since Nov 2025) ingests daily transcripts, people, projects. Gives agents structured context that feels like personal AGI.
30:22 – 38:37
09 · Live Orgo Demo
Nick spins up a cloud computer, triggers Telegram meta-agent to install Hermes inside the VM. Shows 27 customer VMs managed from one agent.
33:53 – 38:37
10 · Cloud vs Mac Mini
Cloud VMs beat local Mac Minis: remote access, sandboxing, instant spin-up/delete, blast radius protection.
38:37 – 43:56
11 · Agents Build Agents
Arm Claude Code/Codex with MCPs: Perplexity, Context7, Exa AI, X MCP. Spawn 5 parallel sub-agents for research synthesis.
43:56 – 45:28
12 · Watchdogs and Observability
Every agent needs: watchdog that auto-restores crashed gateways, and email alerts from agent to operator when cron jobs or skills fail.
45:28 – 47:55
13 · Close
Nick thesis: you and your agent, building other agents for other businesses, is one of the most leveraged positions in 2026.
Lines worth screenshotting.
- The winning offer for a solo AI agent business is unlimited agents, unlimited usage, and unlimited support at $5,000 per month — the unlimited framing removes all friction from the sale.
- Customers do not need as many agents as they think — most businesses need one or two well-configured agents, not five or ten.
- Never mention tokens or credits to clients — the moment they start thinking about usage costs, the magic disappears and trust erodes.
- Sell an AI employee, not an AI agent — frame the delivery as a digital staff member that knows the business and gets smarter every week.
- Going vertical into a specific industry (law firms, agencies, real estate) is what distinguishes you from a commodity Claude Code reseller.
- A customer should be fully onboarded with their first working agent within 48 hours — speed to value is the primary sales retention mechanism.
- Talk in terms of business outcomes and revenue impact, not time saved — time saved is overused and people are now immune to it as a value proposition.
- Hermes agent setup, OpenClaw configuration, and Claude Code skills are genuinely rare capabilities that most businesses would pay $5,000 per month to outsource.
- If you can set up Claude Code, you already have a more valuable skill set than 99% of the businesses you could be selling to.
- The one-person agent agency model scales not by adding headcount but by standardizing the onboarding playbook and increasing client count.
- When an agent breaks for a client who has become dependent on it, the pain is immediate and serious — proactive monitoring before the client notices is the retention moat.
- Legacy industries with no internal AI capability and high operational complexity are the best targets — the value gap between their current state and what you deliver is widest.
The playbook is out in the open.
Steal the framework
Nick gave away the actual mechanics, not just the concept: offer structure, vertical list, exact stack, live setup.
- Frame every product as an AI employee, not a tool. Never say tokens to a customer.
- The unlimited offer ($5K/mo) works because customers over-estimate how many agents they need.
- Diverge-converge: try 3-4 verticals before locking in; let inbound pull you to a sub-niche.
- Agents build agents. Use Claude Code plus MCPs for setup; stop debugging terminals manually.
- Build the observability layer first: watchdog crons plus agent-to-operator email alerts.
- Content is the warm pipeline. Nick got on this podcast from an Instagram reel Greg saw at midnight.
Terms worth knowing.
- AI agent
- Software powered by a large language model that can take actions on a user's behalf — sending emails, browsing the web, running code — instead of just generating text in a chat window.
- Tokens
- The chunks of text that language models read and generate. Most AI providers bill by tokens consumed, which is why usage-based pricing can feel unpredictable to customers.
- Productized service
- A consulting or done-for-you offer packaged with a fixed scope and price so it sells and delivers like a product rather than a custom engagement.
- Wedge
- A narrow, easy-to-sell first offer used to enter a market and earn the right to expand into broader services with the same customer.
- Cold audience
- Prospects who have never heard of you or your offer before the sales conversation. Selling to them is harder than selling to a warm audience that already knows your work.
- Long-horizon task
- An assignment that takes an AI agent many steps and minutes or hours to complete, as opposed to a single quick response. Reliability over long runs is one of the main challenges in agent design.
- Granola
- An AI meeting-notes app that records and summarizes calls. It exposes an MCP server so agents can pull meeting context directly.
- MCP
- Model Context Protocol — an open standard that lets AI agents plug into external tools, apps, and data sources through a uniform interface.
- Trello
- A Kanban-style project management tool where work moves through columns like Backlog, To Do, Doing, and Done. Used here as a customer-facing request board.
- Kanban board
- A visual workflow tool that tracks tasks as cards moving across columns representing their status. Helps cap work-in-progress and prevent scope creep.
- Scope creep
- The gradual expansion of a project beyond its original deliverables, usually without extra payment or timeline. A common failure mode for service businesses.
- Loom
- A screen-recording tool used to send quick async video updates to clients instead of scheduling meetings.
- Calendly
- A scheduling tool that lets prospects book a meeting on your calendar through a shareable link, removing the back-and-forth of finding a time.
- Superhuman
- A premium email client built around keyboard shortcuts and AI assistance, designed for people who process large volumes of email.
- Asana
- A task and project management platform used for internal team workflows and tracking detailed deliverables.
- Claude Code
- Anthropic's coding agent — a terminal- and desktop-based tool where Claude can read, write, and run code against your project. Used here to build other agents.
- OpenAI Codex
- OpenAI's coding agent product, the company's equivalent of Claude Code, that runs as a desktop and CLI application for software tasks.
- Harness
- The orchestration layer that runs an underlying language model in a loop, gives it tools, and manages memory. Lets the operator swap the underlying model without rebuilding the agent.
- Hermes
- An agent harness used in this playbook to deploy customer-facing AI assistants. Model-agnostic and pitched as more reliable and self-evolving than alternatives.
- OpenClaw
- An open-source agent harness that competes with Hermes for running customer-facing AI assistants on top of any underlying model.
Things they pointed at.
20:49toolHermes Agent
21:05toolOpenClaw
23:40toolAgent Mail
39:45toolContext7
40:10toolX MCP
Lines you could clip.
00:10
“People are charging $5,000 a month per customer to build and manage agents for them.”
06:38
“You're selling an AI employee. You're not selling an AI agent.”
33:37
“The answer to all of our problems, Greg, is that more agents is the answer.”
45:48
“You're underestimating how much value that is, and you can really create a lucrative business by yourself.”
11:03
“I can go on a walk, and there is work being done for our business and customers and their agents by my agent.”
Where the conversation goes.
00:00 – 02:54sparseIntro
02:54 – 07:26denseThe Offer: unlimited agents, $5K/mo
07:26 – 14:54denseThe Market: verticals and niche strategy
14:54 – 17:51Getting Customers: content flywheel
17:51 – 20:49Customer-Facing Stack: Granola, Trello, Loom
20:49 – 28:14denseAgent Build Stack: Hermes, Orgo, Composio, Obsidian
25:51 – 28:14denseModel Picks: GPT 5.5, GLM 5.1, Kimi, Opus 4.7
28:14 – 30:22denseObsidian as second brain
30:22 – 38:37denseLive Orgo walkthrough
33:53 – 38:37Cloud computers vs Mac Minis
38:37 – 43:56denseAgents-build-agents MCPs
43:56 – 45:28denseWatchdogs and observability
45:28 – 47:55sparseClose
Word for word.
00:00People are charging $5,000 a month per customer to build and manage agents for them.
00:07This is a startup idea I wish more people would do. The customer doesn't touch tokens or models or any infrastructure.
00:14They just get a digital employee that knows their business and it gets better every single week. In this episode, Nick from Orgo breaks down exactly how to build this business. The tools, the stacks, how to onboard a customer in thirty days, and and how to actually sell to busy executives, agencies, and law firms.
00:34We also share the full implementation playbook. Hermes, Cloud Code, memory layers, skills, all of it.
00:42This type of episode isn't shared anywhere on the Internet. This is the alpha that people keep for themselves. I'm giving it to you for free.
00:51Enjoy the episode, and I can't wait to see what you build.
01:02I couldn't be more excited to have Nick from Orgo back on the pod. Nick, by the end of this episode, what are people gonna get out of it?
01:12Greg,
01:13everyone's gonna learn not only how to run a solopreneur agent business, but every every gap, everything that they're gonna do wrong from the beginning, I'm gonna save them all the time from having to learn from my from from from those mistakes that I made along the way.
01:28And at the end of this video, you're gonna know what what offer to bring to the market, how to get customers, how to fulfill, what's the stack for the agents that you're gonna build out, and,
01:41yeah, I'm excited to just dive right in. So So Nick, this isn't gonna just be like a pie in the sky, I I want a billion dollar idea here. This is how you can take advantage of AI agents to build a business that maybe does a few million dollars a year, but not just not just the idea.
01:59Right? You're gonna actually share all the tactics from a to z so that by the end of this episode, someone could obviously like and comment and subscribe, but, you know, go go and start one of these businesses.
02:13Right? Exactly.
02:15And like, I think the big thing is for everyone who's watching the pod, you're probably already affluent with AI, and you don't give yourself enough credit. And the amazing thing is, like, 99% of the world has, you know, there's like, many people are so behind on AI, and you you may not realize how valuable your skill set is.
02:36Like, oh, if you can set up ClaudeCode, if you can set up Hermes Agent, if you could set up OpenClaw, that's a very valuable skill that a lot of businesses don't have time for, and you can monetize that. So
02:47Alright. I'm intrigued. Let's go.
02:49Alright.
02:50So let me start by I'll share my screen. Okay. So let's just dive right in.
02:57Let's go into the offer. So when you're starting a one person agent business, you need to have you need to remove all the friction for your customers.
03:07Um, they don't wanna think about tokens. They don't wanna think about computer infrastructure, security, you know, breaking it when it you know, fixing it when it breaks, they they just want it to work.
03:19And so the biggest thing is you need to create abundance in your offer. And what I have found in in my own personal success with this is offering unlimited agents, unlimited usage, unlimited monitoring, support, security, ongoing changes, etcetera.
03:36And the key here is you might flinch because you're like, well, how how do how is that even feasibly possible? Well, the way to do this is to realize the point it's not that the customer's gonna actually need unlimited agents.
03:49They're they're not gonna need unlimited tokens, but they might they might think they do. In reality, they might think they need five agents, 10 agents, a 100 agents, when really one, two, maybe three agents goes such a far away, and you can get a lot of juice for squeeze out of just properly taking the time to set, you know, one or two of these up.
04:10And that's where your you know, that's how you're going to essentially, like, control your costs. You're not spending too much money on tokens, and you're gonna charge 5 k a month for this.
04:19And this is the the offer that I've been running, and it's been working really well. And, yeah, like, customers don't really need as many agents as they might think they need, and you're just gonna show them as quickly as possible the magic behind it.
04:34So this is the offer that I've been running off the rip, and I guess I'll I'll read a little bit.
04:41I I wrote some of this stuff down. The big thing here is the point is not that the customer needs infinite agents. They don't need infinite tokens or infinite computers.
04:50Most customers, they just need one, maybe two, maybe three. They just need a seamless experience. Like, that is what you as, you know, as your business, as as the solopreneur agents agency, you're gonna come in, and you're just gonna remove all the friction.
05:06And the minute that things start to break, like, the business owners that you're gonna be selling to, they're gonna become so reliant, so dependent on these agents that if something does start to break, it is very painful for them. And so in this video, like, I wanna make sure that I help you make it very clear on how to prevent those those those small gaps so that when something breaks, you have something in a way to fix it before they even realize it.
05:33And, yeah, if if a customer if they want constant improvements, how do you keep up? How do you fulfill?
05:40We're gonna be going through all of that in this video. Cool.
05:44Yeah. So, I mean, my big takeaway from this is you're selling an AI employee. You're not selling an AI agent.
05:51People need less agents than they actually think that they need. And you wanna think about unlimited You know, you don't wanna use the word tokens basically at all, and you shouldn't really worry too much about usage.
06:09Exactly. Exactly. Because for them, it just it it ruins the magic.
06:14The minute you say like, oh, like, you're gonna be paying for x amount of credits, and then they're always gonna be wondering, like, oh, how many credits do I have left? And then you're gonna be like, oh, and then it's usage based afterwards.
06:24It's like the more clarity, the more simplicity you can create in the offer that it's just straightforward and easy, the faster time to yes, the faster you can get building, and the faster you can just, you know, have a happy customer. So, yeah, that's the offer so far.
06:40And so then the key here is you wanna go vertical. So as always, you wanna clarify, you're not a commodity. You're not just selling, you know, Cloud Code.
06:49You're not just selling ChatGPT. You're you're selling it, like, a vertically specific industry specific agent. You're doing it fast.
06:59It shouldn't take longer than forty eight hours to get up and running with the first agent for your customer, and you need to talk in terms of time, not not time saved, but actually outcomes for the business. So, like, how much revenue can you generate for the business or how much you know, always always business outcomes rather than time saved.
07:18I feel like time saved is a little overused, and and people are kind of immune to that these days. So that's the offer.
07:25It's pretty simple. And I'll just dive into, like, what we're seeing in terms of our own experience of, like, running this offer. I believe that as a one person business, you can sell these agents into industries and really just kinda be not only selling the agents, but also just creating clarity around AI.
07:43Like, I think if you're watching this pod, you understand AI pretty well. You probably have a better understanding than most people, and you might not give yourself enough credit of how valuable that is.
07:55And to be the person who can create clarity around all the noise right now, that alone is valuable. And and then to be able to couple that with the tools to help solve problems in these businesses, it's like you you you're gonna become so so irreplaceable for the business that, yeah, it's really just gonna be, like, you and the agents are gonna be what drives the value.
08:17So I have some industries here. In red, I have health care and finance because I don't think that these are necessarily the best industries to start off in.
08:27They're very high regulatory burdens and and red tape. And so I actually recommend these other industries that we're seeing work really well, which is marketing agencies, law firms, insurance agencies, manufacturers, wholesalers, and real estate agencies.
08:46Um, the reason for these industries that you might notice is that they're relatively, you know, I would say maybe, like, legacy industries, not not not necessarily, like, you know, new fast growing industries, but they want to be fast growing.
09:01They want to adopt AI, and they want they have a lot of pain to be able to use it as a tool to essentially just grow their business. The common pattern with all of them is they want to be a full stack AI company.
09:14Meaning, they want to be, like, fully automated with AI. That's the dream outcome. We're not there yet, um, but you can certainly come in and start solving the problems from the executive level, and then it'll ripple its way throughout the rest of the, uh, the business.
09:34And I'll dive in on some of the common patterns, but how are we feeling so far?
09:39Yeah. I think those are all people businesses.
09:43So there's a lot of people. When you have a lot of people, there's a lot of waste in terms of efficiency and there's ways to automate things.
09:51That's one. Two is those a lot of these companies want to be AI native is another way to say what you're saying, but they don't know how.
10:02They might have pieces of their companies that have become AI native.
10:10They might work with Deloitte. They might work with, you know, different, you know, AI transformation agencies.
10:16But, you know, to assume that these companies are a 100 AI native is insane because Yeah.
10:27They're not. And then the last thing here is these categories are large.
10:34Right? Like law, it's really large. Insurance agency, it's really large.
10:38Manufacturing, it's large. Wholesale, it's large. The key here is once you've identified a category that you want to go after, then you have to figure out what is the subcategory or sub niche that I want to go after.
10:52It's too hard to just focus on wholesalers. But wholesale know, and the way to think about it from a framework perspective is, you know, pick a category and then you can you can do like, you know, real estate agencies in Florida.
11:09So that's like, you know, geography is one way to do it. Or you can pick a specific type of real estate, you know, professional.
11:22So it could be commercial real estate, you know, agencies in Florida. So there's there's different ways to think about how you can niche down, and that's going to be really key here because if you want to create an irresistible offer, you know, a big way to get the attention of someone is to be like, oh my god, this person is really speaking to me.
11:48Exactly. And honestly, like, even a little bit of a, like, some maybe some contrarian advice from my end is, like, you don't I I I a feeling as though you don't have to start super niche from the beginning.
12:01In fact, you can always niche down after trying a marketing industry, trying a law firm, trying all these different industries, seeing what works well for you, where the market pulls you, and then going super vertical. But, um, I really love, like, the concept of, like, it's a design thinking principle of diverge and then converge.
12:20So, like, you know, try try many different things as long as it's not, you know, too for too long because you don't wanna get into this constant cycle of trying something new and, you know, you never get to focus. But once you find the thing that clicks for you, whether it's you're able to resonate with the audience really well or you're just getting pulled into that market more, like, yeah, go super niche, go sub niche, and use that as your wedge to kind of, like, infiltrate the rest of the market.
12:48Yeah. I think that's spot on. So and then as far as the common things that we're seeing.
12:54Right? So within all these industries, what you'll find is the people you're gonna jump on calls with, the people that are going to likely be the decision makers and the the ones purchasing your service or your productized service, these are the executives. These are the decision makers.
13:11And when you abstract on all these industries, the decision maker at the end of the day has very similar problems no matter what the industry is. They have too many emails, too many meetings, too many follow ups, too many open loops.
13:24They have context over so many different projects and places and people to keep track of. And so just out of the gate, if you can anticipate this, you can have something that you put together that, you know, maybe from a template perspective solves a lot of these issues, and then you can cater more specifically into that niche, into that vertical for that industry.
13:45Uh, if it's a if it's a a law firm and you have a partner who wants to buy your services, um, you can have all of these things out of the blocks for your agents that you set up, which I'll show you how to do also in this video. And and and then you could also cater it for that particular industry.
14:04So, oh, yes, we have a agent that does, you know, um, you know, following, uh, following up with people, projects, etcetera, but it also manages your cases. It does demand demand letters for your law firm.
14:19It does all the different things and skills that you would need for for, you know, maybe a matrimonial law firm for instance.
14:27So that's the, uh, abstraction layer on no matter what, you're gonna be solving a lot of executive problems, and then the key is to layer in, uh, vertical specific solutions as well. Okay.
14:42So there's that. That's the market. So we talked about the offer.
14:45We talked about the market. And I have some side things as well about, like, how to get customers.
14:51At the end of the day, I think everyone should make content. I think if you if you if you can jump on a call, this is just a little tidbit.
15:02If you can jump on a call with somebody and they know who you are and what you sell without you having to tell them and they're warm to begin with, that's the ideal position to be. You never wanna be in a position where you're, you know, having a cold call.
15:17You never wanna sell to a cold audience. And, you know, in the beginning, you might have to. So, you know, starting for free even is sometimes worth it just to get case studies and get referrals.
15:27But content is, like, overpowered in 2026, so I I do recommend that.
15:33I mean,
15:34that's how we met, you know. That's how we met. It's like midnight, can't fall asleep.
15:39I'm like doom scrolling Instagram. I see Nick's face pop up showing me, you know, how to use Open Claw, And I was like, this is a guy who has some sauce and I need him to have him on the podcast. So the other thing about creating content is not only is it helpful in terms of getting your face in front of customers or or your offer in front of customers, but it also helps you, you know, get known, get on podcasts, hire the right people.
16:09So there's there's a lot of advantages. And in an AI world, when you can use AI to automate a lot of the research and a lot of the just help it, you know, the editing and things like that,
16:23just do it. Like, I hate to say it, just just do it. You gotta just do it.
16:28It's just like it's the most leveraged thing you can do. It's like content. And then if you think of other, like, leveraged things, it's like, okay, AI or you could also have leverage with talents and software.
16:39But, yeah. It's it's it's incredible.
16:43And I think, like, the trend of 2026 is content is king, and and I'll tell you a little bit of a tidbit as we go into this next segment.
16:53But I don't know about you, Greg. I have been going on walks. And what I'll do is I'll go on a walk, and I'll I'll send off a long horizon task to my agent via Telegram.
17:06I have my own, you know, Hermes agent. It's what I use these days. And it's I'm just like I'm just in awe with what the the world we live in today.
17:16Like, how amazing. I can go on a walk, and there's work being done on for our for our business and on customers and, you know, for their agents by my agent.
17:26And I'm just like, if you extrapolate that over the next six months, twelve months, like, the the most leveraged thing you could do is post a piece of content that reaches a lot of people and then have this robot that helps you fulfill for the thing that you're providing as you go on a walk or right before you go to bed or when you wake up.
17:47It's just it's amazing. It's an incredible world we live in. So, yeah, let's dive into the stack, shall we?
17:53How do we build these things? Okay. So as far as the tools that you might need to fulfill for your service of, you know, providing agents for businesses, First and foremost, I use granola.
18:09I love granola. I use it for every meeting. They have an MCP.
18:13You know, you can give it to your agent, and it just has context over everything. And what I do is these meeting notes from granola, they automatically get synced into requests on Trello.
18:24And so Trello is the customer facing, like, essentially project management Kanban board that I use.
18:33And so, you know, there's a backlog list. There's a to do list. There's a doing list, there's a a done list, and the customer can just simply drag and drop what they want into the to do list for, oh, I want my agent to be connected to my calendar.
18:50I want it to have access to this other platform. I want it to create content for me. And they could just add these requests at one at a time.
18:59And the key here is these agents can, at this point, do so many different things. They could do so many things that you almost need to create you know, prevent scope creep by, you know, limiting one to two requests in under forty eight hours, um, because there's a lot.
19:18And you could do a lot, but you just need to be careful that you don't, you know, end up drowning in a fulfillment nightmare. Um, so that's why Trello is helpful in terms of scoping. Loom is awesome.
19:30Your customers are gonna want you to send them updates, you know, send an update at 2AM, send an update at, you know, different times of the day of you implementing new things for the agent, whether you improve the memory or you improve the the Obsidian vault that it's it's operating off off of.
19:46Loom is awesome. And then I just use, like, Calendly link. Like, I have a horrible funnel, but you can do you can do a lot of you can get a lot of bookings.
19:56Just Calendly link, personal website, drive traffic there, create content. These are, like, pretty much the customer facing tools. And then I have, I don't know about you, Greg.
20:05Do you use do you use Superhuman?
20:07The email tool? Yeah.
20:09I I don't, but people tell me I should. Oh my god. I if you if you have a lot of emails, you're gonna have a lot of emails with customers.
20:17Oh, man. It's Superhuman is amazing.
20:20It has a bunch of shortcuts. I love keyboard shortcuts, and it just flies through emails.
20:26And it's not like it's AI generated. Like, it makes you write the email and you have AI help you, but it's just a very focused focused platform.
20:34And then lastly, Asana. I use Asana for internal facing, so not customer facing. You know, if I wanna keep track of some specifics around details of of what needs to be done.
20:44Yeah. That's the that's the software stack.
20:49Okay. Let's dive into the the agent side of things now. So for building agents, the irony here is if you don't know how to build an agent, please don't worry.
20:58I got you. We're gonna use agents to build agents.
21:04And so Cloud Code, they have a new desktop app. It's awesome.
21:09OpenAI's codex, they have a new desktop app. It's awesome.
21:14And you can actually use these to build the agents for your customer. And as far as what agents to use, you have a couple options. You're not gonna sell Cloud Code to your customer.
21:26You could or Codex. I mean, you could, but I highly recommend using Harmis these days.
21:33I find it to be the most reliable. It allows you to pick any model. The reasoning here is tomorrow, there's gonna be a new model that comes out.
21:42It's gonna be infinitely cheaper, and it's gonna be OPUS 4.7 level intelligence. And it's like, you just wanna have the flexibility to quickly switch whatever the agent that you're running it, whatever model it's running, be able to switch that quickly.
21:56And you don't wanna be married to a platform, married to a tool, married to an infrastructure. So Hermes, I really like.
22:04Have you played around with Hermes at all? I I think I saw some videos.
22:08Yeah. Yeah. I have.
22:11I haven't, you know, quite made the shift yet, but I've done an episode on it with my friend Imran, so go check it out if people are interested in learning about how to set up Hermes. We called it Hermes because we're fancy like that in the episode, and then we got, you know, the team at Hermes, you know, quickly corrected us.
22:33They did? Oh, wow. I I like Hermes.
22:36Yeah. Hermes is a little more fancy. If you if you sell Hermes agents, you can charge 10 k a month.
22:41Exactly. So, yeah, all OpenClaw is commoditized already. You know?
22:45It's it's 5 k a month. It's okay. So you pick your harness here.
22:50So this is, you know, the agent that you'll sell, and you need a place for that agent to live. You can use something like Hostinger.
22:57You can use Orgo. You can use whatever.
23:00Um, I obviously am biased, um, but Orgo is really nice because in one workspace, you can have all your agents.
23:10You have your agent managing their agents, and I'll dive into all of this, uh, and getting set up. And then lastly, you need the tools for the agents.
23:19Some things out of the box that I install for every agent no matter what outside of just giving them a computer and the ability to use it is Composeo.
23:28Have you heard of this company Composeo?
23:31I have, but can you give a one liner for folks who haven't heard of it heard of them?
23:36This company allows you to this connector, they they allow you to have one connector, one MCP essentially, that connects to thousands of other apps, whether it's Gmail, Slack, Notion, what have you.
23:50And with one connection, you can manage you can have access to all the tools that you would need to send an email via Gmail or push something via GitHub or pull a message via Slack.
24:02It's incredible. And it handles the tool, the tool calling, and the authentication, which is huge because security is, like, biggest challenge of setting up these agents.
24:13Like, by far, the biggest time sync is getting authentication set up for the customer because you have what's your username and password for this? And then if you email it, it's, like, not secure.
24:25So then you use something like Composeo, done. So it handles that. And then it handles security in that sense as well.
24:31Everything's managed to their platform. And then it manages it it handles the tool call. So if you have Composeo set up with all the connectors, you can just take that one connector, take it to any agent, and it has all the same connectors.
24:43So I really like this company. I don't have any affiliation, but I love their product. Really great.
24:50Next up is agent mail. This one is I I give every agent an email. It adds a nice personal touch.
24:58So, you know, let's say you're, yeah, you're you're an executive. I give you an agent.
25:04You name it Mia, and Mia needs it's Mia needs her own email. Agent mail allows you to give Mia an email so that she can send and receive emails, and that's really fun because it's it turns into, like, truly, like, a personal assistant.
25:21And then lastly, Obsidian. You have a video on Obsidian.
25:25It did really well. Obsidian, super important because at the end of the day, these agents need context.
25:32And the more context you could provide in a nicely Wiki styled structured format and markdown files for the agent, it will really just thrive in terms of understanding projects, people, things that you're doing, so on and so forth.
25:47So this is the stack. And as far as models, I guess final touch around models, today, by far, the best model to use for something like a Hermes agent or an OpenClaw is GPT 5.5.
26:04It's so efficient with the tool calls. It doesn't eat through tokens like Opus 4.7 from Anthropic does. And and OpenAI is very generous around letting you use your paid plan with with with any model, like, with any harness like like Hermes or or OpenClaw, and then and you just get a lot of usage out of it.
26:27So I recommend 5.5. If you wanna use open source models that are a little more affordable for lighter weight tasks, GLM 5.1 from z AI is, in my experience, like, the best open source model to be using.
26:41Kimi comes in on a at a close second, and these are both more affordable. And then OPUS 4.7, finally, if you have some long horizon coding task, OPUS 4.7 is really great for that.
26:56And you can actually have your agent connect to Cloud Code and be able to do these long coding tasks in Cloud Code and then bring that back to the agent.
27:06So tidbit there.
27:09I want I don't know if you can do this real quick, but could you give a one liner on because people are gonna sorry. Let me take a step back.
27:17People are gonna look at this list and they're gonna be, oh my god. Don't know if I I should use Codec or if I should use Cloud Code, if I should use OpenClub, if I should use Hermes.
27:26Should I use Hostinger, should I use Orgo, should I use this, should I can you go and just quickly you know, what's Nick's stack and, like, with a one liner of why you use that tool over the other tool?
27:38Yeah. Codex because it's more generous and it's simplest and they have the best desktop app.
27:46Hermes because it doesn't break and it's self evolving. OpenCLO is not as self evolving.
27:53Orgo because we give your agent a computer so it can live in the computer. It could operate the computer.
27:59We're not just a headless VPS server in the cloud, and I'll dive in on that. Composio, you need this.
28:07Everyone needs this. Agent mail, everyone needs this. Obsidian, everyone needs this.
28:13And then
28:14That's a hot take, by the way. Obsidian, everyone needs this.
28:18Oh, yeah. Yeah. Just I mean, explain why you should use Obsidian over, say, Notion.
28:27So here's my Obsidian vault. And, Greg, I've been I've been building this vault since November 2025.
28:36Agents. I mean, a what do you how do how do you say?
28:41When something's outdated. Like, super old. 2025.
28:46That's forever ago. You know? So I've been building this vault since twenty twenty five November before OpenCloud, before Hermes, and it has everything about people, projects, everything.
28:59And I'm so crazy. I have limitless microphone.
29:05Even that, daily transcripts get pulled from that into here. This is genuinely a second brain. Like like, people say Obsidian as a second brain, and then, okay, they they show it.
29:17Okay. That's kinda cool. They use it for some research.
29:19No. No. No.
29:19This is a second brain. And when you have something like this, it is quite literally you get to experience what personal AGI might feel like in the next three to six months.
29:33From now, I'm sure everyone will experience it, but I I feel like I'm getting to experience it sooner because I just have such well organized markdown files.
29:42It's incredible. It's incredible. So it just gives your agents context on what it needs to know given any given tasks, and it feels like it just never forgets and it understands you.
29:59And I think that's at the end of the day, like, we just want an agent that understands us and helps us with our business and and just has perfect context over everything we do. So
30:09Enough said.
30:10Yeah. And then 5.5 is the best. I would just use 5.5 to make it easy.
30:15Yeah. GPT five point So as far as that's the stack.
30:21Now in Orgo, we give the agent a a computer to live in. Greg, let me invite you to this. Alright.
30:29So I just invited you, Greg, into this workspace in Orgo, and we're just gonna quickly spin up a computer here. And I'm gonna spin up a computer.
30:38I'm gonna say Greg's computer. I'm gonna launch it, and it launches pretty fast with really fast desktops.
30:48And now that we're in this workspace here in this computer, I can now install the agent inside of it, actually.
30:56So the agent can live inside of here. So if it's OpenClock, if it's Hermes agent in this case, it will live inside of this environment. And the key here is regarding to getting set up, we have an Orgo MCP that is what I use for for setting up agents.
31:15And that little story I told earlier about going on a walk and be able to get work done on a walk, it's because my agent is using a Orgo MCP to connect to my customer's agents that live on Orgo. And so what ends up happening is Orgo is like this workspace where my agent and other agents and myself can all collaborate on these computers where these agents live and get them set up and configured that way.
31:43So here I have a I don't know if can you see my Telegram chat?
31:51Yep. I had my agent last night. Actually, I kicked off a task.
31:55I told it to go ahead and build out some CLI and skills for for for Orgo.
32:02But I'll start a new chat here, and I can actually just tell the agent I'll grab the computer ID from Orgo. So let's grab this computer ID, and I can give this computer ID to the agent.
32:17And I'll say, set this computer on Orgo up computer ID quoted here.
32:28Let's install Hermes agent into the VM.
32:34So the reason why I tell everyone not to get stressed out or scared about setting up these agents is you really just need another agent to set it up.
32:45In my case, I'm using another Hermes agent to set up a Hermes agent. In another case, you could hell, spin up another computer here.
32:54In another case, you can literally install something like Claude code into a VM on Orgo, and you can actually just run Claude code from the terminal here and tell Claude code in natural language, hey.
33:09Let's set up Hermes agent. So just real quick, you know, you got Cloud Code install command for Linux.
33:17You find that real quick, and you just run this in the terminal here, and you would literally install Cloud Code, run it from here, and have it install Hermes into this VM.
33:31So the answer to all of our problems, Greg, is that more agents is the answer. If you're confused on how to set something up, have your agent do it.
33:44And So I'm just gonna install Cloud Code here. It's gonna get that going, and we'll be off to the races.
33:53This will be a dumb question, but why are we doing virtual computers versus doing local computers?
34:02You know, buying Mac minis and and doing that whole thing.
34:06It's actually a very, very good question. And the reason is we want the ability to work on our customers' computers from where we're at.
34:16And if you are using a Mac mini, I can't even imagine the nightmare of having to go in person and, you know, debug something that's, like, at a hardware level or something on the Mac mini breaks or an update or or what have you.
34:32Orgo gives you cloud computers to be able to manage these agents. And with that, you can do so many more amazing things, both from a perspective of just scaling your business, being able to access all these agents on one platform via one connector and have your agent connect to all of them.
34:51That's really, like, the biggest thing. It's just from a fulfillment perspective, it is just, the easiest way.
34:57And then also just a security perspective. These are isolated cloud computers, and you can delete them.
35:04And under a second, you can create a new one. And with that, there's a lot more sandbox environments that you could just protect you and your customers from, like, a blast radius that might otherwise be more dangerous on a personal Mac Mini.
35:20So
35:21And so say I have a 100 customers. Am I creating like, how am I structuring that?
35:27Am I creating, like, separate, like, projects with these computers in it? Like, what is from a best practices perspective in terms of security and just, you know, good UX,
35:39how should people think about setting that up? Yeah. So in this case, it would be exactly like you said.
35:45Like, I if you were if you were a customer, I would just make a workspace for your business, and I would say, you know, this is let's do, like, idea browser.
35:57Right? And we would create this workspace, and each each of your agents would live in this workspace.
36:05And then I'd have other workspaces for other customers, and I'd be able to manage all of that, you know, on on Orgo.
36:14Yeah. One platform.
36:17Cool. Yeah. I think what's also cool about this, just just how visual it is, like, showing this to a customer and being like, I know you think, you know, it's not secure.
36:28You might think it's not secure or you might think, you know, but this is, you know, a visual sandbox environment. Right?
36:35Like, it just it feels like the cell like, you just like you know, you talked about Loom before, but, like, showing looms of this, I think it's just gonna light people up.
36:44Yeah. Exact and also, like, you can also out of the box on Orgo, like, we have this playground mode here.
36:51So, like Mhmm. This is our this is just all of the latest models from, you know, Anthropic and and Kimmy and and ChatGPT. And so as far as a demo, when you tell a customer, like, oh, yeah.
37:03Like, we can we can have an agent, like, operate a computer and and do do things for you, and, essentially, you're describing Hermes agent or you're describing OpenClaw, even then, they they might have a hard time, like, imagining, like, what what what does that look like?
37:16What does that feel like? And so when you just give it a computer, you're able to just give it life, and you could tell them.
37:22Like, you can say, like, hey. Look up what is Idea Browser and search it on Google.
37:29And this actually becomes, like, a really good demo. Like, for for you and your customer, like, to be able to show, look.
37:37Oh, the agent is controlling a computer, and it's doing research, and it's doing real work. And you can just quickly show a demo in Orgo.
37:46It's, like, super cool.
37:49This is cool. Yeah. Like, even me as, like, a cofounder of Idea Browser, I'm, like, looking at this, and I'm like, yes.
37:56Yeah. Yeah. It's awesome.
37:58And and we have some I'm I'm I'm making Jordan some agents. I don't know if he told you. I'm I'm making him some idea browser agents, and he's, uh, he's using them.
38:07He's stress testing them. Um, but everyone needs agents.
38:11So it's, um, yeah, it's really cool. And then as far as the telegram set up here, like, you can see my agent is literally using Orgo MCP to get this Greg's computer set up right now, and it's installing Hermes agent.
38:28Part of this yeah. These some of these things take time. It's running a long process.
38:33It might take five, ten minutes, so not to bore you by sharing that. But the concept of use agents to set up other agents, it's very real.
38:42And I can also dive into practices around that and how to make sure that your agent knows how to set up other agents.
38:52Yeah.
38:53Does that sound good? Let's do that. That sounds great.
38:57So to have your agent have context of how to set up other agents, it actually needs I wish I would have added them here. A few MCPs go really far away.
39:08One of them is the Perplexity MCP.
39:13With Perplexity, you can give your Claude code or codex real real up to date knowledge on things like Hermes.
39:21And so the key here is, like, you always just wanna have any sort of setup process, initialization process grounded in real context of what is the docs for setting up Hermes agents today and what how do I connect my Hermes agent to iMessage?
39:38If you have something like Perplexity, you give your agent the ability to see how to do that and be able to set it up perfectly. Exa AI is another great MCP tool for, like, real time web search.
39:52Another big one actually is Context seven. This one is awesome for getting up to date docs from, like, GitHub from, like, Hermes agents GitHub so they they can see specifically the docs of how to get set up.
40:06You just need some sort of context layer to let to loop in the best practices and up to date docs for setting up these agents. And it kind of, like, one final recommendation would be the XMCP.
40:18So Twitter released their own MCP, and I find so many amazing setups on Twitter for Open Call and Hermes agents that there's many times I just wanna use that context for setting up an agent for a given task.
40:32And you could actually use this, give this to your Cloud Code or give this to your codex and have it use this context to help help you set things up.
40:42Or you could use all of them too. So
40:45I mean, is there any downside to using all of them?
40:48No. I I use all of them. And so, like, maybe even in here when you look at my telegram, you might notice oh, I guess it's it's I didn't ask it to to pull in up to date context.
41:00Mine has skills already built in place to be able to, like, set these things up just because I do it so much. But in general, like, the more the merrier, like, context is key.
41:11And I like to, like, have sub agents spawn. I'll tell Codex, like, hey or Cloud Code, hey.
41:19Spawn five sub agents. One sub agent for Perplexity, one for XO, one for Context seven, one for Fire Crawl, one for XMCP.
41:29Because I like to pull from different resources, and then those all come back to the main agent, and we get the best practices.
41:37So that's how I do it.
41:40Cool.
41:42Let's see. Okay. So this here is the I should have ran this in the terminal below, but I'll go ahead.
41:50I'll I'll just run it in this terminal. And what we're gonna do here is do that, run Claude.
41:58We could just like spin up multiple terminals.
42:01Yeah. You can oh, wait.
42:05Let me see. Okay. I think I just copy this here.
42:09So okay. Oh, command not found.
42:15Okay. I'll I'll debug this later. This is also why this is mainly why I use the Orgo MCP.
42:23I'm just like, I let my agent do all the work. Actually, you could also come here into the playground and say, install Cloud Code into this computer and just have our agent do it because I I don't I don't wanna debug what's going on in the terminal right now.
42:40So then just have this agent do it and do it that way. But, yeah, once you have it set up from here, like, I can now ask my Telegram agent.
42:52So I'll I'll stop this here. You could just imagine, you know, when that's done, your agent's set up, and I'll start a new chat. And I'll ask, like, how many Orgo VMs do I have in my workspaces?
43:08And my OrgoClaw is able to actually manage all of my customers' agents from, you know, just this one agent, and it can upgrade, fix things on the fly, you know, and all from one spot anywhere I'm at.
43:29If I if I get a email from a customer that something broke, we can just send off an agent send off a message to OrgoClaw and have it go fix it.
43:42Boom. You can see here, 27 Orgo VMs across your workspaces, all 27 shows running. And then it dives in onto all the different customers and all their agents.
43:55So last point that I wanna make around getting these things set up is the watch docs.
44:03So the gateways are what make these agents connect to a platform like Telegram or a platform like WhatsApp.
44:13And sometimes these gateways crash. OpenClaw has a lot of gateway issues in my experience.
44:19Hermes is a lot better. And so a key here is you want to make sure that you set up a watchdog.
44:27You could literally just tell your agent, set up a watchdog for whenever a gateway crashes that it auto restores it. Um, that's super important just from, you know, reliability perspective.
44:40A second thing is you wanna make sure that you have some layer of observability or alerts. So I have agents email me. If I set up your agent, your Mia agent, and Mia has an email, Mia's Mia emails me from her email when her cron job breaks or her skill failed or something happened, and I'm alerted about it.
45:08And I can go in and then debug it and fix it, which is super valuable because once again, for your customer, you don't want them to have to worry about, like, doing all this themselves.
45:19So make it as simple and easy as possible, handle everything tip to tail, and, yeah, I think I guess, like, the big takeaway here is it is hard to set up Cloud Code even.
45:34Like, people are like, Cloud Code is is gonna kill OpenClaw. Cloud Code is gonna kill Hermes agent.
45:39And in a general sense, it's getting better at doing a lot of these general things. But to be able to go in and create a specific agent for a specific industry and person and have it tailored to their workflow, it's like you're underestimating how much value that is, and you can really create a lucrative business by yourself.
46:00You and your agent building other agents for other businesses. Yeah. And I think it's an amazing time to be a solopreneur for this.
46:09You can and you will. So Nick, thank you for sharing the playbook for how to build a one person agent led business, sharing how to actually do it in such a clear way.
46:25I love chatting with you because you're you give the sauce, but you also explain it super clearly. Nick is a criminally under followed account on on social media.
46:37You know, he's getting some some followers, I think he can be I think he needs to be bigger, so I'll include links for where to find Nick in the show notes, in the description. And Nick, I'll see you in a few weeks in San Francisco, and let's have a let's have some coffee and have a good time.
46:54Thank you, Greg. Always a pleasure.
46:57Thank you so much, and I I hope to yeah. We're gonna see you soon.
47:01We're gonna get some coffee. We're gonna we're gonna do some sipping time.
47:05We're gonna do some IRL sipping time, which is my favorite. It's there's there's nothing like it, you know? Like, I actually have been trying to cut down on my, like, Zoom meetings and stuff like that.
47:18It's just there's not there it is. There's nothing like being in person, sharing ideas, sipping, and and figuring out what what we can be building in in in a time like this because there's so much, and sometimes the hardest part is figuring out the right idea, the right time, the right playbook, the right steps, the right order.
47:42And this has been helpful, Nick, and and definitely got my great creative juices flowing, so I'm sure others are very thankful as well. So thank you, Nick, and I will see you next time. Thank you, Greg.
47:53Talk soon.
The Hook
The bait, then the rug-pull.
The number lands before the explanation: five thousand dollars a month, per customer. Greg Isenberg opens with that line from a voice-over, then hands the floor to Nick Vasilescu, co-founder of Orgo and a practitioner who is actually running this business, to walk through every decision from offer to observability.
