AI Business Marketplace

We build the businesses. You buy the code and run them up*

skip to the listings

← indexlocal plumbing businesses

Plumber Agency

Every town has plumbers making six figures with a website from 2009. This agency sells them the fix — and your Claude does the building.

price
$597 one-time
status
field-tested · held by our review
Inquire about Plumber Agency
the product you sell — client site — a live render of the real file in the box, not a mockup. open it ↗

We are not selling this one yet, and the reason is us. It passed its field test — the machine runs, and the full report is on this page. It has not passed our own pre-sale review, which asks the harder question: does anything in this box put a claim nobody verified into your mouth, or onto a real client's website? Ours found some. Until that comes back clean, this kit does not go on sale at any price.

Inquiring puts you first in line, costs nothing and commits you to nothing.

licence Non-exclusive licence: you are buying the machine, not a territory. Put your own name on it.

the demos

in the box

7 parts, 98 files — yours to rename, reprice and rewrite.

Look inside the box ↗ — the real file tree, and the open files free to read in full before you pay us anything.

  1. 01

    Your agency site — two designs

    Your own storefront, finished and written, in two looks — services, how it works, pricing, FAQ and a wired contact form. Fill one config file, run one command, drag the folder onto free hosting — about five minutes. Not a theme, and not something you have to write.

  2. 02

    The client website

    The thing you actually sell: the site your Claude builds for a plumbing client, shipped as two complete designs — plus a claims checklist that stops you publishing anything the client hasn't confirmed.

  3. 03

    AI operator layer

    The whole system is built Claude-ready. Plug in your own Claude and it runs the machine. Setup is a terminal window, a few pasted commands and a free Python download — about 25 minutes, with a guide that shows you what each one prints when it works.

  4. 04

    Local lead scraper

    Point it at your city. It pulls the businesses in your niche that free map data knows about — names, numbers, sites — into a call-ready list, ranked hottest-first. Map coverage varies hard by metro, and the field test on this page shows you both a good city and a bare one.

  5. 05

    The call script

    Opener, discovery, price anchor, close — plus the objections answered. Written for the phone, not tested on a real prospect: nobody has sold with this kit yet, and that is in the field test. Your Claude preps each call from the lead's own details.

  6. 06

    The prompt pack

    The exact prompts to run the business with your AI: outreach, follow-up, fulfillment, invoicing.

  7. 07

    Growth playbook

    Week one, month one, and how to scale past us. You'll always know the next move.

Built for: Someone willing to make phone calls to local business owners. No agency experience needed — the system knows the moves.

day one

  1. 01Connect your Claude with the included guide — a terminal window, a few pasted commands and a free Python download, about 25 minutes. The kit calls the Python step the single most common place people get stuck, so it walks it through screen by screen.
  2. 02Run the scraper on your city. It builds a ranked list of the plumbers the map knows about, and tells you how many of them arrived with a phone number you can actually dial.
  3. 03Your Claude preps your call list and drafts openers from the script.
  4. 04Start dialing. First offer is simple and cheap to fulfill: their new website, built by your AI.
  5. 05Close one → the fulfillment prompts walk your Claude through delivering it same-week.

what it costs

$597 one-time, no subscription

Market rate for a local business website build runs $500–$1,500. Market rates, not promises — we have never sold one, and what you close is entirely on you.

the licence, and why it costs what it costs

We would rather say this on the page than let you find it out later. Each kit is for sale to whoever wants it, not to one buyer per city. That is deliberate, and it is the honest explanation for the price: a one-of-one business costs thousands because it is sold once. This is a system built to be sold many times, so it costs what a system costs. We make no territory or exclusivity promise of any kind — we could not see it, let alone enforce it. Your edge is your name, your city and the calls you make.

So rebrand it. Your agency name, your domain, your prices, your wording, your voice. The kits are built for that: one config file carries your details through every document, the storefront copy is yours to rewrite, and the demo businesses in the box are invented placeholders meant to be replaced. The files are the same for everyone who buys. The part that is yours is the part you write.

And the part nobody says out loud: if you don't rebrand, the storefront you put up is the storefront they put up. Someone in the next town over can buy this kit tomorrow, change nothing, and ship the identical site with a different phone number on it — same headline, same sections, same words. That is not a risk we can insure you against; it is arithmetic. Rebranding is not decoration, it is the whole of what makes the business yours.

read the full licence →

Inquire about Plumber Agency

30-day refund policy · you also need your own Claude account; that is not our money and we take no cut.

the long version

What it is, in fullread it →close ↑

Plumbers make real money and run on word of mouth — most have weak websites, no follow-up, and missed calls that cost them jobs. This agency sells them exactly that fix: AI-built sites, missed-call textback, review engines. You own the agency; your Claude does the heavy lifting.

Built for: Someone willing to make phone calls to local business owners. No agency experience needed — the system knows the moves.

field test

Field-tested 2026-08-01, re-audited 2026-08-03: a fresh Claude ran day one from the kit files alone, and an outside reviewer then caught our own lead numbers overstating what a buyer can actually dial.

The full report — everything that broke, in our own wordsread it →close ↑

A fresh Claude with no prior context ran this kit end to end. Then, on 2026-08-03, an outside reviewer bought it, followed the instructions literally against four real cities, and caught our own field test flattering itself.

We had been counting rows, not phone calls. Re-measured on the number that matters — leads you can actually ring today — our own reference city read 41 rows and 5 phone numbers. Tampa: 6 rows, 3 numbers. Toledo, Ohio: 1 row, 0 numbers. The Houston run never finished; the free map servers returned 504s. Meanwhile the playbook was telling that same buyer to make 25 dials the next morning.

What we changed: the scraper now headlines dialable leads, automatically widens to the surrounding metro when your city is short, and no longer throws away a whole multi-city run when one city times out. Same command, same afternoon, same reference city: 41 rows / 5 dialable → 107 rows / 17 dialable.

Then we introduced a bug fixing it. With the metro widener on, neighbouring towns overlap, so the same plumber arrived twice under two different city labels — the CEO would have rung him twice. The merge now matches on the map's own record ID first, then the phone number, then the name within two miles. Verified on the worst case, Phoenix + Gilbert: 63 duplicate groups became 4, and all four survivors are genuine separate branches.

17 is not 40, which is what the kit asks for before a full week of calling. Closing that gap is neighbouring towns, the phone-lookup list, or the optional Google Places source — and the scraper now tells you which one your run needs instead of making you guess. Toledo is a genuine OpenStreetMap coverage desert: 3 businesses in a 30-mile box, 1 with a phone number. No engineering fixes that. Run your own metro before you buy.

One more thing that audit found, in our own files: three income claims — a “pays for itself” line and two cousins of it — that should never have shipped. Deleted, and the whole kit re-scanned.

The tester was an AI, not a customer. Nobody — not us, not a buyer — has sold a website with this kit yet. That is the part most sellers leave out, so we put it at the bottom of our best page.

read the kit's own field-test file →