AI cannot answer "what's in my food?" today because the data is split across 15+ databases that don't talk to each other.
| Source | Has | Missing |
|---|---|---|
| USDA FoodData Central | Nutrition for 380K+ foods | No prices, no stores, no real-time |
| Open Food Facts | 4M+ barcodes, crowdsourced | Sparse US coverage, stale data |
| openFDA | All US food recalls since 2004 | No products, no nutrition |
| Kroger API | Real-time prices, stock, aisle location | Kroger-only, no MCP |
| BLS / USDA AMS | Commodity & retail food prices | No products, monthly lag |
| Instacart | 85K+ retailers, fused catalog | Centralized, not sovereign, closed |
Aggregate every free public source, add retailer API data, maintain origin provenance on every record, serve unified MCP tools to any AI.
| Source | Records | Updates | License |
|---|---|---|---|
| Kroger Product API | 6,346 products | Real-time | Free (OAuth, 10K/day) |
| USDA FoodData Central | 593 products | Quarterly | Public domain (CC0) |
| Open Food Facts | 1,045 products | Continuous | Open Database License |
| openFDA Recalls | 5,000 recalls | Weekly | Public domain |
| USDA AMS Market News | 20 prices | Weekly | Public domain |
| BLS CPI Average Prices | 39 prices | Monthly | Public domain |
A physical grocery store has one layer — the box on the shelf. The AI shelf has nine. Every layer is a surface where value flows — to the consumer, from the brand, verified by the data.
Nutrition, ingredients, allergens, country of origin, certifications, SNAP eligibility. Provable, from authoritative sources.
The box we "deleted." Brand story, product positioning, sustainability claims, comparison messaging, new product launches. Brands pay to syndicate.
Customer ratings, purchase patterns, repeat purchase rate, trending products. Community-generated, brand-amplified.
Professional endorsement. "Featured in Ina Garten's Caesar Salad." Same romaine used by Sweetgreen. 47 recipes across NYT Cooking.
Social media endorsements, viral recipe trends, dietitian recommendations, fitness content. Paid and organic.
AI-computed diet compatibility, health goal matching, drug interactions, glycemic impact. Generated from nutrition data.
Farm identity, harvest date, cold chain verification, miles traveled, carbon footprint. The FSMA 204 / data wallet layer. What nobody else has.
Promotions, digital coupons, Ibotta cashback, bundle deals, loyalty points, price comparison across stores.
The only layer that flows FROM the consumer. Active medications (GLP-1), recent surgery (gallbladder), chronic conditions, per-person allergies, life stage. Lives in your data wallet, not the retailer's cloud.
| Dimension | Google Search | Food Intel |
|---|---|---|
| Data unit | Web page | Product at a specific store on a specific shelf |
| Price | Maybe mentioned on a page somewhere | $2.99 right now at Kroger On the Rhine |
| Availability | Unknown | In stock: HIGH |
| Location | Store address | Aisle: VEGETABLES, Bay 16, Shelf 3, Right side |
| Nutrition | Generic article about romaine | 20 cal, 1g protein, per this specific product |
| Safety | News article about a recall | Real-time recall check against FDA database |
| Comparison | 10 blue links to blog posts | Side-by-side nutrition + price for 5 products |
| The query Google can't answer | "Find me every SNAP-eligible, gluten-free, sub-$4 protein source in stock at my store sorted by protein-per-dollar" | |
Any AI can discover and use these tools via /.well-known/mcp.json
| Tool | What It Does |
|---|---|
food_search | Search products by name, brand, or UPC barcode across all sources |
food_nutrition | Full nutrition panel + ingredients + allergens for a specific product |
food_compare | Side-by-side nutrition comparison of 2-5 products |
food_recall_search | Search FDA food recalls by product, company, or reason |
food_recall_check | Check if a product or brand has active recalls |
food_commodity_prices | Agricultural commodity prices (terminal market + farm gate) |
food_retail_prices | BLS average retail prices for staple grocery items |
food_ingredients_lookup | Search products by ingredient or allergen |
food_data_sources | Full transparency: all data sources, record counts, last update times |
food_new_products | Recently added products — shows the new product discovery lag |
food_supply_chain | FSMA 204-style traceability: farm-to-shelf journey |
The grocery industry has a 5-layer data infrastructure with no cryptographic spine. Every layer runs on trust-by-contract.
| Layer | Who | What | Rootz Opportunity |
|---|---|---|---|
| Identifiers | GS1 | GTIN / UPC / Digital Link | Signed product identity wallets |
| Syndication | 1WorldSync, Syndigo, Salsify | Brand-authored product content | Verified, signed brand claims |
| Retailers | Kroger, Walmart | Store-specific price + stock | We already pull from Kroger API |
| Aggregators | Instacart Universal Catalog | Cross-retailer fused catalog | Sovereign per-product identity dissolves this moat |
| Consumer | Apps, AI agents | What the shopper sees | Food Intel — this is what we built |
.well-known/ai endpoints belong — signed, agent-readable product identity. The food-product analog of Origin's company profiles.
Every endpoint returns origin provenance. Every response tells you exactly where the data came from.
GET /.well-known/mcp.json — MCP discovery
POST /mcp/tool — Execute any tool
GET /api/search?q=romaine+lettuce — Quick search
GET /api/sources — Data source transparency
GET /ai-context — AI orientation & skills