Reserve a share of a local farm animal or other farm goods: priced, paid, weighed, settled and delivered across one API, a mobile app and four web portals.
BrownBox is a marketplace where households reserve a share of a live animal or other goods from nearby farms. BrownBox sets the price, takes payment, settles the final weighed price and pays the farms. It is built as three apps: an Express and MongoDB API, a Next.js site with portals for operations, farms, drivers and pickup stations, and an Expo customer app.
I wrote the entire API and the mobile app, and co-built the web portals with a teammate. That covered the pricing engine, weigh-in settlement, farm payouts, Stripe payments, QR handoff, shipping, realtime chat and push notifications, with 858 API tests running against a real in-memory database.
I built it with a spec-driven agentic workflow: project rules and skills for Claude Code in each repository, plus system specs and decision logs guiding every phase. BrownBox is pre-launch and will go live at brownbox.farm.
Captured from a local build of all three apps running the repo's demo data (fictional farms and customers). Product photos are public-domain stand-ins.
Core functionality and capabilities that make this project stand out
Key Technologies I Used on This Project
• Turning BrownBox from a broker into the seller mid-build, across three apps, in six phases.
• Keeping customer money invisible to farms, enforced by per-audience serializers and tests.
• Concurrency without transactions: atomic guarded writes for capacity, stock and payouts.
• Keeping card payments honest when an amount changes after the payment request is sent.
• Keeping AI-written code consistent across three repositories over four months.
• Spec-first agentic development (a knowledge base, numbered decisions, as-built notes) keeps AI agents accurate on a large codebase.
• Clients never send prices: integer cents on the server are the single source of truth.
• Provider interfaces with console fallbacks let the whole stack run and be tested without keys.
• Tests against a real in-memory MongoDB catch what mocks hide.
Deeper dives into how parts of it were built.
How a knowledge base, a numbered decision log and as-built notes kept an AI coding agent accurate while building BrownBox's API, web portals and mobile app, with 858 tests to keep it honest.
This project showcases practical development skills and problem-solving experience. Explore the features, review the implementation, or get in touch to discuss similar work.
