Real food data, real scores, real vibes — across 1.87M products.
Full dev docs and SDK references — API reference, SDKs, MCP integration guide — ship at docs.mintberry.ai.
real food data · real vibes
What people buy isn't just nutrition. Food Signals wraps our real food data layer with vibe scores, taste profiles, pairings, and occasions — so the product page you ship feels like something people actually want.
Carbonated water, high fructose corn syrup, caramel color, phosphoric acid, natural flavors, caffeine.
no parsing · no flags · client must split + classify
Typed schema. Built-in vibes. ~620× cheaper.$499 / $0.80 ≈ 620×.
Nutritionix Starter: ~$499/mo floor (seat-based, you pay it regardless of call volume) ÷ 1,000 calls = $0.499/call. Food Signals Build: $0.0008/call × 1,000 = $0.80. FatSecret Premier isn't priced publicly; their Basic tier is free with attribution but US-only data and capped at 5K calls/day.
Public per-call pricing. No seats, no annual minimums, no sales call. The data the other two gate behind $499/mo or price-on-application — typed, scored, and cited from $0.0008 per call.
quickstart
Three steps. No enterprise red tape, no kickoff call, no quote in your inbox a week later.
One field, one email. Production keys are pk_live_ — usually back to you the same day.
Or a name. Hit /v1/signals/food. Get a strict JSON response back.
Pure HTTP works. Native SDKs for TS, Python, Go. MCP server pprmt/food-signals for agents.
for teams shipping ai agents
If you're shipping an AI surface that touches food — shopping agents, wellness coaches, kitchen helpers, voice — here's what plugging us in actually changes.
Direct access to the product catalog — real ingredients, macros, micronutrients, additives, and allergens for every food.
While other agents burn tens of thousands of tokens on fighting search engines for partial answers, yours just calls a function.
Calibration instead of hallucinated confidence.
Every score, every ingredient, every macro links to a citable upstream — so when a user asks "why?", the agent has a real answer.
NOVA, Nutri-Score, and EWG-style frameworks — three peer-reviewed traditions, every score recomputable from public inputs.
When your agent says NOVA 4, it points to the paper.
Not to a proprietary 58/100 nobody can defend or compare.
Multi-source, categorized with Google Taxonomy, continuously refreshed, and composite-scored by people with the expertise to do it right.
Your roadmap stays focused on what's actually your product.
MCP-native, strict JSON, stable schemas.
Your agent calls Food Signals like it calls any other tool — same shape in, same shape out, every time.
Posted pricing on the same page as the docs.
No NDA, no procurement, no kickoff call.
Wire it up in an afternoon, ship it next week.
posted pricing
Free until you ship. Linear when you do.
For nights, weekends, the side project that turns into something.
For production traffic, paid users, and the moment things actually work.
$0.0004 / req over 5M / moFor high-volume production with custom retention and a dedicated channel.
open methodology · v1 · published spec
Three measures, three traditions of nutrition science — peer-reviewed, citable, recomputable. Returned as one JSON response with source links.
Industrial-processing class. Class 4 = ultra-processed. The most predictive single signal in nutrition science.
read methodology →Front-of-pack nutritional grade. Negatives (sugar, salt, fat) net against positives (fiber, protein, fruits).
read methodology →Composite score per product plus per-ingredient flags. Every flag links to FDA position, EFSA position, and the underlying study.
read methodology →No NDA, no procurement, no kickoff call. Full dev docs at docs.mintberry.ai.