Skip to main contentSkip to footer
    Back to Case Studies
    Pet BreederBuilt with Claude Code
    Claude CodeWeb PlatformIn-House Product

    Pet Breeder Directory, Built With Claude Code

    How one founder and Claude Code built a breeder, shelter and breed directory since March 2026. This page tells the problem, the idea and each stage of the build.

    Project Preview

    Live website preview - petbreederdirectory.com

    1,419
    Commits from March 22 to October 25, 2026
    1,249
    Commits written with Claude as co-author
    605
    Lines of project rules in CLAUDE.md
    240
    Unit test files, plus 27 browser test files

    Rules that Claude reads first

    A 605 line CLAUDE.md and 8 rule files tell each session about the stack, the data and the billing. They also list the claims that the copy must never make.

    Skills for repeated work

    10 project skills package the work that repeats: a bug fix, a release, an investigation, a retrospective, a litter sync and the blog article pipeline.

    Subagents with one job each

    12 subagents review code, plan work, check security and check TypeScript. 8 of them form a support pipeline that can diagnose a ticket against the code.

    Tests as the guard rail

    240 unit test files and 27 Playwright files run before a change ships. Some tests guard words, not code: one fails the build if copy calls a free tier verified.

    Small reviewed changes

    80 pull requests and 82 database migrations, each migration applied by hand. Claude writes the change, and a person decides when it goes to production.

    Search work in the open

    In October 2026 the site had 1 page indexed of about 316,000 known URLs. The fix was a rule set that Claude follows: index the core pages first, then open more in stages.

    The problem

    A person who wants a puppy or a kitten from a breeder must do a lot of research. The data exists, but it is in many places: kennel club registries, breed clubs, shelter databases and thousands of breeder websites. It is in many languages and many formats.

    A buyer who searches for one breed near one town usually finds a marketplace that lists animals like products, or an old forum thread. Breeders and shelters have a different problem. They have no single place where they can show their work and keep their information correct.

    The budget was a second problem. A directory needs years to earn search traffic. A stack with monthly fees for each user, each row or each seat can stop the project before the traffic arrives.

    The idea

    Pet Breeder Directory is a Frenchy Digital product, founded by Chris Machetto. The idea was to collect the public breeder data in one searchable place, and to give each breeder and shelter a profile that they can claim and manage.

    The second part of the idea was the method. A small team cannot write a directory of this size by hand in a few months. We used Claude Code, the coding agent from Anthropic, as the main engineer, and we kept a person in charge of each decision. We wanted to learn what this method can build, and what it needs from the people who use it.

    What the directory is, and is not

    • Aggregated, not vetted: The listings come from public sources. A small number are claimed by their owners or verified after a review. The site never says that breeders are screened as a group.
    • Majority European by coverage: The largest country counts are the Czech Republic, France, the United States and the United Kingdom. The United States and Canada are the sales focus, not the main coverage.
    • Early on the paid side: Billing works, but this page reports what the code does. It does not report traction that the product has not earned yet.

    How it was built, month by month

    The first commit is from March 22, 2026, and the most recent is from October 25, 2026. That is about eight months of work. The commit counts below come from the git history.

    • March 2026, 98 commits: Claude Code wrote the Next.js scaffold, the database schema, the design system and the homepage on the first day, from a sequence of numbered prompts. In the same week it added 9 international breeder scrapers, the breed health data and the breed quiz.
    • April 2026, 243 commits: The busiest month of the early build. Claude added the puppy search page, the breeder data enrichment pipeline and the admin email campaign manager. It also added lead capture for Google sign in and the Sentry error setup.
    • May and June 2026, 81 commits: Billing became complete: cancellation emails, retries for failed payments, refunds by country and file delivery for digital products.
    • July and August 2026, 114 commits: Claude added search and filters on the directory pages, paid boosts at the top of results and the Claimed badge. A scheduled job now confirms each breeder website again.
    • September 2026, 70 commits: Breeder profiles began to list available and sold dogs with prices. The breed quiz became a lead form, and the unsubscribe flow got its own page.
    • October 2026, 813 commits: The largest month. About 300 commits were search and indexing work. Claude also built the support inbox, a social media pipeline and a check for every blog article title and description.

    How Claude Code was used

    The project memory

    Claude Code reads a file named CLAUDE.md at the start of each session. In this project the file has 605 lines. It holds the stack, the deploy steps, an audit table of production numbers and the rules for the copy. 8 more rule files cover the API, sign in, code style, the database, email, git and tests.

    The file is not documentation for people only. It is the reason that session 500 knows what session 50 learned. When a mistake costs time, the fix goes into the code, and the lesson goes into CLAUDE.md or into a test.

    Skills and subagents

    10 project skills

    A skill is a written procedure that Claude runs on request. This project has skills for a bug fix, a release, an investigation, a product review, a retrospective, the litter sync and the blog articles.

    12 subagents

    A subagent is a separate Claude with one job and a limited tool set. Four review code, plans, security and TypeScript. Eight form a support pipeline: triage, an orchestrator with four checks, specialists for interface, sign in, data and settings bugs, and a reporter.

    A person stays in charge

    Claude writes the change and the tests, and the tests must pass before a change ships
    Database migrations are generated SQL, and a person applies each one by hand, never on deploy
    Decisions about what Google may index, prices and claims belong to the owner, and CLAUDE.md says so in writing
    Production numbers in the rules carry a date, and the rules tell Claude to query them again before it uses them

    What we learned

    Tests can guard words

    The site has a reviewed verification tier and a free website check. The copy must never merge the two claims. A unit test fails the build if any string calls the free tier verified.

    More pages is not more traffic

    The code can make about 16.9 million breed by city pages. In October 2026, Google had indexed 1 page of about 316,000 known URLs. The owner chose to index the core pages first, and the rules now open the rest in stages.

    Memory is the main tool

    Speed came from Claude. Quality came from the rules that the team wrote down after each problem. A project without that file starts from zero in each session.

    Repeat work becomes a skill

    Work that repeats became a skill with fixed steps. That makes the result the same each time, and it makes each step easy to check.

    Where it stands

    Reported honestly

    The directory is live at petbreederdirectory.com. The numbers on this page come from the git history and from the dated audit table in the project rules. We do not publish revenue or subscriber counts here, because they are small. For the architecture of the platform, read the Pet Breeder Directory platform case study.

    Describe what a feature does, never promise content that does not exist.

    — Pet Breeder Directory project rules

    What This Project Does Well

    A working directory on the first day. The scaffold, the database schema and the homepage came from written prompts on March 22, 2026
    About 252,000 breeder records imported from public sources, with 9 international scrapers in the first week
    583 breed pages across 8 species groups, and a breed by city page space of about 16.9 million URLs
    Stripe billing, custom sign in with two-factor codes, a breeder portal, a shelter portal and a lifecycle email engine, all written with Claude Code
    A project memory that grows: each costly mistake became a rule in CLAUDE.md or a test, so the next session does not repeat it
    Every claim on this page comes from the repository: the git history, the rule files and the audit table in CLAUDE.md

    Technologies Used

    Claude Code (CLAUDE.md, rule files, skills, subagents, headless runs), Next.js 15, React 19, TypeScript, Tailwind CSS 4, Drizzle ORM, PostgreSQL 16 with PostGIS, Meilisearch, MinIO, Stripe, Resend and React Email, Sentry, Vitest, Playwright, Bun, Docker, Coolify, Cloudflare

    Ready to Build Something Similar?

    Let's discuss how we can help transform your business with a custom digital solution.