# Cheapest Grocery Basket
Per-call grocery basket optimization for AI agents. Answers the decision an agent cannot compute itself: where should this entire shopping list be bought today, at real local store prices. Every price is read from the retailer's own storefront for a specific store and carries the timestamp it was read. Package sizes are normalized to a common unit so a gallon and a half gallon compare honestly. Paid via x402 (USDC on Base).

## Endpoints
- GET /find?q=milk&zip=30501  ($0.005) — one item across every collected store: price, sale, size, per-unit price, stock, link
- GET /basket?items=milk,eggs,bread&zip=30501  ($0.01) — the whole list priced at each store, with missing items called out
- GET /cheapest?items=milk,eggs,bread&zip=30501  ($0.02) — cheapest single store AND the cross-store optimum, with savings, stops and substitutes
- POST /mcp — MCP server (find_product, price_basket, cheapest_basket); tools/call is priced identically to its HTTP route
- GET /sources — how prices are collected and matched, and what is not collected

## Free
GET /health, GET /llms.txt, GET /sources, GET /server.json, and MCP initialize + ping + tools/list.

## Honest limits
Coverage is per collected ZIP and per collected product; anything else returns an explicit not-covered error and is NOT charged. Delivery fees, order minimums and loyalty-card prices are not collected yet and are returned as null rather than estimated. Shelf prices, promotional prices and stock are live.