Andrew Mattes lives outside Philadelphia. Wife, two young kids, a full-time job he genuinely likes, and part ownership of a brewery that he calls his foray into small business ownership. Somewhere in the gaps, he's trying to buy a company.
He started under a year ago knowing nothing about any of it.
Today he has an AI agent running on a Mac mini at home, connected to his Searcher OS account, that triages the listings that came in overnight, checks them against his buy box, and writes his follow-up questions into the CRM before he picks up the phone with a broker.
He's not a developer. He taught himself by asking Claude how to do things and then doing them.
I put him on camera for 15 minutes to walk through the whole setup.
Square zero
Worth sitting with where he started, because it's the part most searchers will recognize. His first move was the Acquiring Minds podcast, just listening to people who had already done it.
“I pretty much started from square zero. [...] I grew up near and around some small business owners, but in my mind, that was just something you either passed down to your kids one day or closed up shop at the end of a great run.”
That's the gap, stated better than I've heard it stated. Plenty of people grow up around small businesses and never once consider that you can walk up and buy one.
The podcasts led to more research, which led to Paul Lajoie's community, which is where he and I met. Then came BizBuySell, a wide net, and no real idea what he was looking for. (Everyone does this phase. It's fine. It just doesn't survive contact with nights and weekends for very long.)
Under a year later his buy box is tight: Philadelphia, New Jersey, and the Jersey shore, fairly agnostic on industry, with home services and home health as the two he's honed in on.
What MCP actually is
Here's Andrew's explanation, which is better than the one I usually give:
“We say MCP, that's just the connection, essentially the way that my instance of Claude can speak to Searcher OS and has the ability to read all of the data that's on that side.”
That's the whole concept. MCP is a connection. Your AI assistant talks to your Searcher OS account directly instead of you copying financials into a chat window one deal at a time.
It runs both directions. He queries the deal flow sitting in Searcher OS, and he updates his broker notes and moves his own tasks without leaving the window he's already working in.
The part he mentions almost in passing is context. His agent carries the history of what he's looked at, why he passed on certain deals, why things weren't the right fit. The agent holds the memory of his search instead of just the current screen.
It also pushes back. The morning we talked, it flagged that he had two listings sitting in the CIM analyzer for weeks and asked what the story was. Kill those, move on.
Days to seconds
If you only have 30 seconds, this is the part to read.
“Even just before this call, I was able to triage the new deals that came in overnight, compare them to existing things in the deal funnel, log notes directly to my CRM, and then tee up time later next week to give a broker a call. Those are things that would have taken me days, weeks of manual time. That's just a matter of seconds now.”
Look at the list of jobs, not the number. Triage overnight listings. Compare them against what's already in the funnel. Write the notes to the CRM. Line up the broker calls.
That's the actual weekly grind of an acquisition search, and it's most of the reason searchers with full-time jobs quit around month 4.
He described the old version too: pull up this CIM, what did it say, now normalize that against the other one. Anyone who has done that by hand knows how much of a week it eats.
His rule for what gets handed off is simple enough to steal.
“Anytime I'm doing something that takes me more than 10 minutes of manual effort, why am I doing this? How can I offload this to Claude and how can I make this more efficient?”
He taught himself by asking
I asked how someone with a full-time job and no coding background built any of this.
“I think AI is probably the most underrated education tool in the world, because everything I've done has just been asking Claude how do I do this and then implementing what it says.”
Ask, implement, ask again. That's the loop, and it's the same loop whether you're wiring up a connector or learning a new industry.
He was also clear that the CRM is what he came for first, well before any of the AI, and he still uses the platform on its own. Scrolling deal synopses and saving things to the funnel. The email inbox integration, which he singled out. His words. Anyone who's done this knows how chaotic it gets when brokers change firms and you're working the same broker across two different deals.
The connector sits on top of a system that was already keeping him organized. That order matters. (An agent pointed at a mess gives you a faster mess.)
Where to start
His advice is smaller than you'd expect, which is most of why I believe it. “Just think one thing that you'd like to offload and just try,” he said. “It's really just a matter of taking a baby step today.”
Come with a problem first, then reverse engineer the solution. Pick one repeatable thing you don't like doing.
He's since connected Gmail, Square, and QuickBooks for the brewery. He can run a P&L review in seconds and send himself a monthly look-back on a schedule he picks. None of that was the plan on day 1. It came out of asking what he wasn't using yet.
The thing he deliberately protects is conversations. He'd rather spend his hours talking to brokers, owners, and other searchers, because that's where the introductions come from and it's the piece AI can't do for him. The admin is what he handed off.
If you're already using Searcher OS and you've never switched the connector on, it lives in Settings, under LLM Connect. For Claude and ChatGPT it's one URL and a browser login. No config files, no API key, no terminal.
Set up your MCP connection, or read the step-by-step connection guide first.
Thanks to Andrew for the 15 minutes.
Andrew Mattes is a self-funded searcher in Philadelphia, PA.