AI Development Lab

Your quotes take
three days.
Ours take
ninety seconds.

Live AI quoting engines for home service contractors. Any trade that still prices jobs by hand.

We plug straight into the supplier APIs you already buy from, and pull real material pricing at the moment a customer asks. Your labor rates and your markups are loaded in as a custom matrix, so the number that comes out is the number you would have written yourself. It just arrives while the homeowner is still on the page.

What actually happens in those ninety seconds.

Three pieces of infrastructure, doing the job an estimator does, at the speed an API does it.

01

Supplier API Integration

A direct connection into the distributors you already buy from. Ferguson, ABC Supply, and the regional houses that have an endpoint. Material pricing is pulled live at the moment of the quote, so the engine is never working off a spreadsheet somebody forgot to update in March.

Live pricing, not cached
02

Custom Labor Rate Matrix

Your rates, your crew sizes, your job classes, your markup on each. We build the matrix from how you actually bid, then the engine applies it the same way every time. Your margin is a rule in the system rather than a decision somebody makes at 9pm when they are tired.

Your margin, enforced
03

Automated Job Intake

The homeowner answers the questions your estimator would have asked. Scope, access, measurements, photos. The engine scores the job and prices it. Anything genuinely unusual is flagged for a human instead of being guessed at.

Qualified before you drive

Two ways to price a job.

One of them costs you fifteen hours a week and you are already doing it.

The Old Way

Drive out, measure up, hope they call back.

  • Fifteen hours a week in a truck, quoting jobs for free.
  • Most of those visits were tire-kickers and you found out after the drive.
  • The phone rings while you are under a sink, so the lead goes to whoever picked up.
  • Quotes go out three days late, priced off a supplier list from last quarter.
  • Margin gets guessed at the end of a long day, then shaved to win the job.
  • A chatbot on the site that collects a name and an email, then asks them to wait for a call.
The Engine

Priced while they are still reading the page.

  • Material cost pulled live from the supplier API at the second of the request.
  • Labor and markup applied from your matrix, identically, on every job.
  • Intake runs day and night, so nothing depends on who is free to answer.
  • High-ticket work priced on the spot instead of sitting in a callback queue.
  • Margin calculated by the system, so it holds at 11pm the same as 9am.
  • Unusual jobs flagged to a human. The engine knows the edge of its own competence.

We were infrastructure people before this.

Nobody should hand their pricing to a lab that cannot show its work. Every tool below is public, MIT licensed, and running. Open any of them and read the source before you talk to us.

Open source ledger MIT licensed
System Role Stack State
mcp-doctor Health check, lint and security audit for any MCP server, from the command line Python, zero deps Flagship
memory-mcp Persistent searchable memory for agents over MCP, atomic JSON storage, weighted keyword search MCP server, 25 tests, CI Shipped
promptlint Linter for prompt and agent instruction files, catching token budget overruns before they ship Python, zero deps Shipped
claude-mcp-starter-kit MCP configs, tested prompts and a memory system that turn Claude into an autonomous operator Claude, MCP Maintained
Core Product

The AI Quoting Engine

One engine, wired to your suppliers and loaded with your numbers. It runs on your site, prices the job while the customer is still reading, and hands your team a qualified job instead of a name and a phone number.

From 8,000 USD per install
$ quote --job=repipe --sqft=1850
-> supplier api: 34 line items priced
-> labor matrix: 2 crew, 3.5 days
-> markup applied: material 22%, labor 18%
-> margin check: passed
Quote 14,880 USD. Sent in 82 seconds.

The figures in that terminal are an illustration of the output format, not a live quote. Real numbers come from your suppliers and your matrix.

You already sell to contractors. Now you have the product.

This is the price authority for the engine. Thirty percent of every install and thirty percent of the monthly, from a floor of 8,000 USD. You keep the client relationship you already own. We do the build, the integration, the deployment and the support. Every month you wait, a competitor learns the engine exists.

Marketing Agency Owners

You have the contractors.
You are missing the product.

You already run ads, sites and lead gen for home service contractors. Plumbers, roofers, painters, whoever is on your roster. You already know their real problem is not traffic. It is that a good lead sits in an inbox for three days while somebody finds time to price it.

We built the thing that fixes it. You put your name on it.

We do the supplier integration, the labor matrix and the support. You place it with clients who already trust you, at a price they will pay because it pays for itself. Thirty percent of every install, and thirty percent of the monthly. No developer on your payroll and no product to maintain.

Install floor From $8,000
Your share of the monthly 30%
Your share of the install 30%

You place it once and you earn on it every month it runs. The shops that hear about this from a competitor first will not be calling you. Ten clients on your books is a book of recurring revenue, not ten one-off invoices. You are not reselling a subscription. You are placing an asset.

Pre-launch

The math a contractor already lives with.

A shop owner or their estimator spends the better part of two days a week turning leads into priced quotes. Measuring, pulling supplier prices, checking margins, retyping it into an email. That time has a cost, and it is paid every week whether the job closes or not.

The engine does that pricing the moment the lead lands. The install starts from 8,000 USD, once. Set that against the number on the right and the decision stops being about software.

Hours quoting by hand, weekly 15 hrs
Loaded cost per hour $90
Spent pricing jobs each year $70,200

The install is a fraction of one year of that. Every week the old way stays in place is another $1,350 spent to do slower what the engine does on arrival.

Agencies: book a demo

See the engine run on a real trade, get the partner terms and the margin sheet. Name and email, nothing else. No call required and no sequence of nine emails.

Run a shop yourself?

If you price jobs by hand and want to see the engine on your own trade, send it over. Same deal, name and email only, and a straight answer on whether your suppliers have an endpoint we can use.

Tell us the job you quote most. We will tell you if the engine can price it.

A straight answer, usually the same day. If your suppliers have no usable endpoint, we will say that instead of selling you a build.