The AI Team That Runs Searcher OS

When I pulled 90 days of US listings out of the Searcher OS database, the broker market came out to 2,473 active firms putting up about 769 new listings a day on a rolling average. And 796 of those firms, almost a third of the directory, list exactly one deal a quarter.

That's the market a searcher is fighting. It's wide, it's slow, and most of the inventory you want is scattered across two- and three-person shops that aren't sitting by the inbox.

I run Searcher OS solo against that market. The way I keep it running is a small AI team, and I want to be honest about both what they do and what they don't, because most of what gets written about AI agents skips the part where they break.

The org chart

There are 6 of them, and each one is pointed at a single number: grow ARR for Searcher OS.

Bob is my CEO. He sets the weekly OKRs and runs the board where work gets assigned. Bill is my CTO. He watches the error logs overnight, stages bug fixes in a branch, and waits for me to merge (I keep the keys to production). Bridget tracks customer health and flags churn risk before it lands. Riker runs content and SEO. Howard is the knowledge layer, everything I know about buying businesses, synthesized into something searchable.

Then there's Marcus. He's the one I get asked about most, so let me be careful here.

What Marcus actually is

Marcus runs the admin layer of deal sourcing on my own setup: initial outreach, NDAs, CIM requests, first-pass review. The stuff nobody enjoys.

He is not a feature inside the trial. He's a proof of concept I run on my own machine, and I'm describing him so you understand the operating model, not so you go looking for a button. Right now Marcus is how I test whether the most tedious part of sourcing can be pushed into the background. So far I think the answer is yes, with a human still making every call on which businesses are worth buying.

I'd rather underclaim that than oversell it. The honest version: Marcus is me proving, on my own time, the same bet the product is built on.

Why this matters to a searcher

From the people I talked to while building Searcher OS, chasing brokers for documents that were promised and never arrived ran a few hours a week, every week. That's a tax on every active searcher, against a market where a third of the firms surface one deal a quarter and the good inventory hides in the long tail.

The product is the part of that bet a searcher actually gets. Python agents scrape 350+ broker sites daily and drop the listings into one feed, 67,000+ of them. You set a buy box (price range, industries, states, EBITDA) and the feed narrows to deals that fit. The pipeline tracks each one from Interested through NDA, CIM review, LOI, and diligence, and it tells you which deals have gone stale. The CIM analysis reads the PDF brokers send so you're not the one combing it line by line.

None of that is the flashy part. It's the boring plumbing that maps a 2,473-firm market down to the handful of deals worth your weekend, so you spend your hours reading deals instead of begging for them.

The bet, stated plainly

Searcher OS is profitable and growing, used by self-funded searchers, micro-PE operators, and a few family offices. I built it with AI coding tools and zero Python skills, and I run it the same way I built it: one disciplined person plus a few well-scoped agents doing the work that used to need a team.

That's the same thing I'm selling. The org chart above is the actual operating model, and the product is what happens when you point that model at the broker market on a searcher's behalf.

If you're running a search and you're tired of being your own listing aggregator, the feed and the pipeline are what I'd want you to test. Searcher OS has a free plan, no card required. Point your buy box at the market and see what the engine surfaces.

That's the part I'd actually want a searcher to judge me on.