Nick from Orgo built a Grok Bot clone in one weekend.
36 hours. One chat thread. Loops running overnight while he slept.
Then he turned it into a Grok Bot business: custom AI agent teams he sells to normal companies for $25,000 up front, plus an optional $3,000 a month retainer.
No funding and no team. One founder, a coding agent, and an offer specific enough that busy owners say yes.
Greg Isenberg had him on the Startup Ideas podcast to walk through the whole thing: the demo, the stack, the pricing, the first-client playbook. I pulled out every number so you don't have to scrub through 49 minutes.
What Nick Built (and Why 75K People Watched)
Nick is a co-founder of Orgo, the company that gives AI agents their own cloud computers.
He didn't like how the existing agent platforms worked. So he built his own over a weekend and called it Bops, short for business ops.
Every bot in Bops gets its own phone number, its own email address, and its own computer with four screens.
In the episode he calls his chief of staff bot, Boppy, on a real phone number. Mid-recording. He asks it to pull up a list of companies to sell to, check the S&P 500, and watch Hacker News for relevant posts. Then he hangs up and you watch the bot do all of it on its own screens.
He posted a demo of this on Twitter and got 75K views. His words: it almost did as well as Orgo's YC launch video, it took two minutes to make, and he's eating cereal in it.
Then Orgo's customers saw the video and started asking for their own version. Right now he's building a custom-branded Bops for an insurance agency.
I've covered people running their own companies on Grok Bot agent teams before. Nick went the other way: he builds agent teams for other businesses and charges like a staffing agency.
Why Custom Beats the Big Platforms
You might ask: why would a dental clinic pay $25K for a Grok Bot clone when Grok Bot costs $200 a month?
Because the owner of that clinic is never going to set it up.
The big platforms (Grok Bot, OpenAI's Dots, Meta's Muse) are not in the business of building niche setups for insurance agencies or dental clinics. And the owners are too busy running the actual business to learn a new platform. They don't want a tool. They want their problem solved.
Nick's bigger thesis: the era of dumb SaaS is ending. The price of software is trending toward zero, and the money moves to the layer on top: agents, plus human intelligence.
Greg puts a number on the niche premium. Grok Bot charges $200 a month as a generic tool. A version purpose-built for one industry, pre-loaded with the right agents? He estimates 10 to 20x the pricing power, before you even add seat-based pricing.
One tactical gem from Nick's own setup: he builds one agent per platform. A Twitter agent. An email agent. A Skool agent for his community. And a LinkedIn agent that connects with prospects, messages them after the connect, and books the call straight into his calendar. That's a full lead nurturing sequence running while he does something else.
By the time Nick joins the call, the brief is already written.
The 36-Hour Build, Step by Step
Nick built Bops in a single long-running thread inside T3 Code with Opus 5.5. He never connected it to GitHub during the build. One thread, 36 hours, loops running overnight.
His words: "I built this like an idiot." It still shipped.
The process behind it is the repeatable part:
Step 1: Research before anything. His literal first message: build a version of Grok Bot, Meta's Muse, or OpenAI's Dots, and research them first so you understand. Prime the model on what you're even making.
Step 2: Design before code. He designs everything in Paper (think Figma for the AI era) through its MCP. His trick is diverge, then converge: ask for three or four variations of a feature, then combine the best parts of each.
Step 3: Pick the stack. For Bops: Orgo for each bot's computer, Agent Mail for email, Agent Phone for the phone numbers, Composio for connectors. One API connection to Composio and every bot gets Slack, logins, the works.
Step 4: Make the agent read the docs. For every feature, Nick tells his coding agent to read the docs and the API references. Every single time.
Step 5: Screenshot everything. Bug? CleanShot it, send it, say "fix this". See a UI element you like in Grok Bot? Screenshot it, and the agent implements it in your app. Greg adds: keep a swipe file of references for exactly this.
Step 6: Ship fast. Host on Orgo, Fly, or Vercel. Post the demo. Nick's line that stuck with me: you're 323 prompts away from building your anything.
The $25,000 Offer, Broken Down
This is where most builders undercharge, so pay attention to the structure.
The offer to an SMB: $25,000 up front for a custom agent system the owner keeps. Maybe half up front, half at the end. Then, after a 30-day implementation, an optional downsell: $3,000 a month to maintain it.
The sales math writes itself. Nick's pitch to a clinic owner: how much do you pay your front desk person to handle email, billing, bookings, and appointments? The answer is $50,000 to $70,000 a year. For one person.
Then he points at the screen: that's one of the 15 bots I'm handing you today. And you own the system.
Ownership is the psychological unlock. Business owners hate feeling held by the neck on a retainer for AI employees. Sell them the system outright, and the $3K a month becomes a choice instead of a leash.
Greg stacks three sweeteners on top to make it irresistible:
→ Include a fully redesigned website for free. Most SMB sites are dated, and with the same tools you can design something gorgeous in a day. Owners know a website alone can run $25K.
→ Add a money-back guarantee tied to their actual bottleneck: if you can't respond to leads faster, the retainer comes back.
→ Reframe the retainer. $3,000 a month is roughly a $12-an-hour hire in the owner's mental model. That's the comparison they're already making.
And because building is now cheap, you can keep bundling: Nick suggests free lead magnets, like a braces cost calculator built into an orthodontist's site.
How to Land the First Client (Without Picking a Niche First)
The common failure mode: pick a niche off a YouTube video, build the website and the positioning, then realize a week later you don't care about that niche at all. Start over. Repeat.
Nick's fix is to diverge before you converge, with industries too.
Go through your phone. Find the 50 people you know who run a business, or who sit high up in one. Call them. Tell them you're building a whole team of agents plus apps plus a website, all in one, for businesses. Ask if that sounds interesting.
A few will say yes. Do the first one free, in exchange for the case study.
That's how Nick ended up in the promotional products industry. Swag and merch for companies. He didn't know it existed until a warm contact pulled him in. Big, and barely touched by AI.
Then: don't scale yet. Sit with that first client. Ask them to share their screen while they do their manual work, like creating an invoice or building a product presentation. Watch. The agents you need to build become obvious.
Greg's referral twist on the free build: tell them you normally charge $50,000 a year for this service. It's free, on one condition. If it works, they text three colleagues in the same space with an intro. Sweeten it with a $5,000 credit toward their retainer.
Run it that way and you never sell cold. Every deal comes through someone who already vouched for you.
Should You Copy This?
The objection you're thinking: if Nick built it in a weekend, where's the moat?
Greg's answer: that's what they want you to think. The moat is the value you create, plus the fact that the client is buying you, not just the software. Businesses pick the person they want to work with for years, not just the software.
Two more data points before you dismiss it.
First, Nick brought up an Andreessen Horowitz stat: about 2% of the US pays for AI tools. If you can build with this stuff at all, you're underestimating how rare that skill still is. That 2% is heading to 4, 8, 20.
Second, Nick has done this climb in stages. He built his first money-making app after watching Greg's episode with Mickey teaching Cursor back in summer 2024. Then he sold single Hermes agents to businesses at $5K a month, the same model I broke down in the solo AI agent business playbook. Custom Grok Bot systems are just the next rung: agents and apps together, solving the whole gap in one business.
And he open-sourced Bops. The repo is on GitHub (linked in the video description). Point your coding agent at it, fork it, customize it for an industry, and go sell it.
The 36 hours are optional. The 50 phone calls are not.
FAQ
What is a Grok Bot business?
It's building a custom version of a personal agent platform like Grok Bot, pre-loaded with AI agents for one industry, and selling it to businesses in that niche. Each agent gets its own phone number, email, and computer. Nick from Orgo sells these systems to SMBs for $25,000 up front plus an optional $3,000 a month retainer.
How much does it cost to start?
Almost nothing beyond subscriptions. Nick built Bops in 36 hours using T3 Code (free, works with a Claude subscription), Orgo for the computers, Agent Mail for email, Agent Phone for numbers, and Composio for connectors. He open-sourced the result, so you can fork the repo instead of building from scratch.
Do you need to be a developer?
No. Nick describes the whole build in plain English prompts: research the category, design in Paper, pick the stack, tell the agent to read the docs, and screenshot bugs back to it. The skill that matters more is understanding one industry's pain points well enough to pre-build the right agents.
How much can you charge for custom AI agents?
Nick's SMB offer is $25,000 up front plus $3,000 a month, anchored against a $50,000 to $70,000 a year hire. If you productize for a whole industry instead, Greg estimates 10 to 20x the pricing power of Grok Bot's $200 a month, with seat-based pricing on top.
Every week I interview bootstrapped founders doing $100K to $10M a year and pull the playbooks out of them, just like this one.
Listen to the Profitable Founder Podcast →