We Built a Soil Valuation Portal Powered by Live Fertilizer Prices
Agri-Green Farm Management helps farmers understand the dollar value locked in their soil. Before we came in, their valuation process looked like this: soil sample data arrived as spreadsheets, a person converted nutrient levels to dollar figures by hand using whatever fertilizer prices they had on file, and the final report got assembled in a Word document. Every new client ate hours of calculation work.
The problem wasn't that the math was hard. It was that doing it manually, at any scale, doesn't work.
What We Built
We built a Soil Asset Recovery portal with three user tiers: public visitors hit a lead funnel with instant estimates, farmers get a full project dashboard tracking valuations over time, and Agri-Green admins can manage clients and run valuations at scale.
The stack is Next.js, TypeScript, Supabase, and Tailwind with multi-tier authentication via role-based access control. Reports export as branded PDFs with nutrient breakdowns, year-over-year comparisons, and recovery recommendations.
The Part That Makes It Useful
Static fertilizer pricing tables break fast. Prices shift with markets, and a valuation built on stale numbers misleads the farmer. We solved this with a custom MCP server connected to the FRED economic database.
FRED (the Federal Reserve Economic Data API) publishes real fertilizer price indices. Our MCP server queries that data on demand and feeds it directly into the valuation engine. When a report runs, it's calculating against current pricing, not whatever someone typed into a spreadsheet three months ago.
The MCP server pattern here is worth noting. Instead of baking price data into the application or running a cron job to refresh a table, we built a lightweight server that Claude can call as a tool. The AI component runs the valuation logic, pulls live pricing as needed, and returns structured output the portal can render immediately.
The Result
What used to take hours of manual work now generates in seconds. A farmer submits soil sample data, the portal runs nutrient analysis against live fertilizer pricing, and a full branded report is ready for download. The admin dashboard lets the Agri-Green team manage every client account and kick off valuations without touching a spreadsheet.
The lead funnel on the public side gives prospective clients an instant rough estimate before they ever talk to a human, which changes how the first sales conversation goes.
The Bigger Pattern
This project is a good example of how we approach AI builds. The interesting part wasn't the PDF generation or the dashboard UI. It was identifying where stale data was killing accuracy and solving that with a live data layer the AI could actually use.
Most manual processes in small business have a data freshness problem underneath them. Fix that, then automate around it. That's the approach.