> **AI agent?** nutrition-intelligence is a nutrition API for agents (MCP, REST, CLI). **Get a key:** [https://nutrition.prospre.io/register.md](https://nutrition.prospre.io/register.md), a single POST, and the key works immediately. **How to use it:** [https://nutrition.prospre.io/docs.md](https://nutrition.prospre.io/docs.md). Every page on this site is also Markdown: this one is [https://nutrition.prospre.io/index.md](https://nutrition.prospre.io/index.md).

# nutrition-intelligence: proof-backed nutrition facts for AI agents

**Any agent can know exactly what's in anything a person eats.**

nutrition-intelligence is a nutrition database built for AI agents. It holds facts about 581,759 generic foods, branded products and dietary supplements. Every value carries its unit, its basis (per 100 g, per 100 ml or per serving), the strength of its proof and a link to its source. Unknown values are reported as unknown, never as zero.

Agents use it through MCP, REST or a CLI. People can browse and filter everything on this site.

## What's in it

- 25,800 generic foods
- 434,000 branded products
- 121,959 dietary supplements
- Reference intakes, upper limits and the labelling rules of 8 regimes (US, CA, EU, UK, AU/NZ, JP, CN, CODEX)

## Start as an agent

1. Register: `POST https://nutrition.prospre.io/v1/register {"agent_name": "<your name>"}` returns a key that works immediately. Details: [https://nutrition.prospre.io/register.md](https://nutrition.prospre.io/register.md)
2. Call it. MCP config:

```json
{
  "mcpServers": {
    "nutrition-intelligence": {
      "type": "http",
      "url": "https://nutrition.prospre.io/mcp",
      "headers": { "Authorization": "Bearer <your key>" }
    }
  }
}
```

Or REST:

```sh
curl -s https://nutrition.prospre.io/v1/resolve \
  -H "Authorization: Bearer <your key>" -H "content-type: application/json" \
  -d '{"texts":["2 large eggs scrambled","1 cup cooked white rice"],"fields":["energy","protein"]}'
```

Full instructions: [https://nutrition.prospre.io/docs.md](https://nutrition.prospre.io/docs.md).

## Try a filtered search

- [Foods highest in protein per 100 kcal](https://nutrition.prospre.io/search.md?kind=food&sort=protein&per=100kcal)
- [Branded products with 20 g+ protein and under 5 g sugar](https://nutrition.prospre.io/search.md?kind=product&min.protein=20&max.sugars=5&sort=protein)
- [Generic foods highest in fibre](https://nutrition.prospre.io/search.md?kind=food&sort=fibre)
- [Breakfast cereals with the least sugar](https://nutrition.prospre.io/search.md?q=cereal&kind=product&sort=sugars&order=asc)
- [Lowest sodium breads with lab or label proof](https://nutrition.prospre.io/search.md?q=bread&sort=sodium&order=asc&rank=2)
- [Energy drinks with the fewest calories](https://nutrition.prospre.io/search.md?q=energy+drink&sort=energy&order=asc)

Search URLs take filters as query parameters, so an agent can build one directly. See [the search page](https://nutrition.prospre.io/search.md) for the parameters. Without a key, search filters and sorts on energy, protein, fat, total carbohydrate, fibre, sugars, sodium. With a key, every nutrient works, such as magnesium, potassium or caffeine.

## Sources

- USDA FoodData Central, Foundation Foods
- USDA FoodData Central, SR Legacy
- USDA FoodData Central, FNDDS survey foods
- USDA FoodData Central, Branded Foods (manufacturer label data)
- NIH Dietary Supplement Label Database
- Health Canada Licensed Natural Health Products Database
- ANSES-CIQUAL French food composition table
- ADEME AGRIBALYSE life-cycle inventory database

## How values are proven

Every value points to its source, and the strongest proof wins:

- Rank 1: The brand's or restaurant's own published data, or government laboratory analysis.
- Rank 2: Photos of the physical package tied by a barcode, or label data held by a regulator.
- Rank 3: A single panel photo, government calculated/imputed values, or a government dataset whose values are compiled from other sources or modelled (e.g. a life-cycle model).
- Rank 4: A retailer or marketplace listing.
- Rank 5: A front-of-pack or partial photo.

A correction fixes the past: when a value turns out to be wrong, the corrected value replaces it everywhere. Agents that use the data help keep it right by flagging suspicious values and submitting better proof.

## Terms

Free for personal, non-commercial use. Commercial use needs a paid plan. No scraping, bulk extraction, redistribution or model training without a licence. Calls are logged for 7 days, without your identity, to improve the service.

---

[Home](https://nutrition.prospre.io/index.md) · [Search](https://nutrition.prospre.io/search.md) · [Register for a key](https://nutrition.prospre.io/register.md) · [Agent instructions](https://nutrition.prospre.io/docs.md) · [Data sources](https://nutrition.prospre.io/sources.md) · [llms.txt](https://nutrition.prospre.io/llms.txt) · [OpenAPI](https://nutrition.prospre.io/openapi.json)