Heathy — nutrition data (MCP server)
MCP serverDev toolsThis integration adds a nutrition dataset to your AI, covering 917 plant foods and 39 Ayurvedic herbs. Once added, your AI can look up nutrient information for these foods and herbs, with values reconciled across 11 national datasets.
Use Heathy — nutrition data (MCP server) in Claude, ChatGPT or Ahel Desktop
Free. Sign in, add Heathy — nutrition data (MCP server) and connect your AI. About a minute.
Also: Claude Code · Cursor · Codex
Then ask your AI: use the lookup food tool from Heathy
No key or account is asked for. Not verified by us yet: Ahel has not run a call through it.
Details
Available today. Use it from your connected AI after setup.
No key or account is asked for. Not verified by us yet: Ahel has not run a call through it.
After adding it, ask your AI about the nutrition of any plant food or Ayurvedic herb in the dataset.
What your AI can do with it
- Look up nutrition facts for 917 plant foods
- Find nutrition details for 39 Ayurvedic herbs
- Get nutrient values reconciled across 11 national datasets
- Answer questions about what nutrients are in plant foods and herbs
Install Heathy — nutrition data (MCP server)
The server’s own address, for the clients that take one directly. Or connect Ahel once and every client you use reads it from one address, with the account kept on Ahel rather than in each client’s config.
Claude Code
claude mcp add --transport http --scope user heathy-nutrition-data-mcp-server 'https://mcp.heathy.org/mcp'Run it once in your project, then open /mcp to approve any sign-in the server asks for.
Claude Desktop
https://mcp.heathy.org/mcpAdd a custom connector in Settings, paste this address, and approve the sign-in.
Cursor
cursor://anysphere.cursor-deeplink/mcp/install?name=heathy-nutrition-data-mcp-server&config=eyJ1cmwiOiJodHRwczovL21jcC5oZWF0aHkub3JnL21jcCJ9Open the link and Cursor adds the server at that address.
ChatGPT
https://mcp.heathy.org/mcpIn Settings, enable Developer mode, create an MCP app, and paste this address. Your plan and workspace must allow custom apps.
Codex
codex mcp add heathy-nutrition-data-mcp-server --url 'https://mcp.heathy.org/mcp'Run it once, then sign in with codex mcp login heathy-nutrition-data-mcp-server if the server asks for an account.
From the project's README
As published by chandata/heathy-mcp in README.md.
Nutrition for more than 1,100 whole plant foods and their flours and 39 Ayurvedic herbs, drawn from 30 food-composition tables covering 26 countries plus West Africa, the Pacific and FAO's global biodiversity table — including IFCT 2017 (India). It also answers which GB/EU and US nutrition claims the published data supports for a raw whole food or an unfortified flour.
Endpoint: https://mcp.heathy.org/mcp — Streamable HTTP (JSON-RPC 2.0)
Website: https://heathy.org · Contact: hello@heathy.org
Operated by CCT Technologies Ltd (England & Wales, company no. 06337577)
Free while in preview.
This repository contains no server code — only the manifest, these notes and a
Dockerfile. The server is hosted; there is nothing to install. TheDockerfileonly bridges stdio to the hosted endpoint, for directories that inspect servers in a container.
What makes it different
Most nutrition APIs hand back a single number from a single national table. This one returns the number and its provenance:
- the median across the datasets that have the food,
- how many sources contributed and which ones,
- the min–max spread between them, and every source's own value.
So an agent can cite a figure instead of asserting it, and can tell when the sources
disagree. Guava vitamin C, for example, comes back as 222 mg — the median of
13 tables, ranging 68–273 — with 20 national values listed side by side.
That spread is real: soil, cultivar and analytical method differ between countries,
and it is worth knowing before stating a figure as fact.
Indian foods are covered from IFCT 2017, so questions about karela, horse gram or millets are answered from Indian data rather than a US substitute.
How far the cross-checking goes
Be aware of coverage when you use the data:
- Fewer than half the foods are held by two or more tables — up to 26 for common produce.
- More than half come from a single table. Their
sourcescount is1; there is nothing to cross-check them against. Usemin_sourcesonnutrient_rankingto keep these out of a ranking; they are then listed beside it. - Some tables' licences allow their figures to be shown but not averaged. Those
values appear under
by_sourcebut are not folded into the median, which is whysourcescan be lower than the number of values listed.
Connecting
Add to any MCP client that supports remote servers:
{
"mcpServers": {
"heathy": {
"type": "streamable-http",
"url": "https://mcp.heathy.org/mcp"
}
}
}
No API key, no signup. A service card describing the server is at https://mcp.heathy.org/.well-known/mcp/server-card.json.
Tools
lookup_food
Full per-100 g nutrient profile for a whole plant food or flour, with provenance.
| Argument | Type | |
|---|---|---|
food | string | required — slug, common name, or regional alias |
nutrients | string[] | optional — limit to specific nutrients rather than the full profile |
The full profile covers about 130 nutrients, plus a research tier marked
tier: "research" (thin coverage, or a restatement of another nutrient).
What comes back when the name is not an exact match:
- An everyday name ("potato", "kela", "besan") opens the food most national tables
carry. When the match was loose, the answer says so in
matched_from, with the other foods it could have meant. - Everyday nutrient names ("fiber", "omega 3", "calories", "vit c") are understood. A
name that is not is listed in
unrecognised_nutrients; the rest of the answer is returned. - A cooked or prepared form ("lentils boiled", "brown rice cooked") is refused, and the reply names the unprepared food that is held. The unprepared figures are not returned in its place: per 100 g they are a different food.
- A food that is not held — animal foods, foods made from plants (tofu, pasta, sugar), refined wheat flours — gets a reply saying which of those it is.
search_foods
Resolve a name to a slug, including Hindi, Sanskrit and botanical aliases. Returns ranked candidates for a plain-English or local name.
| Argument | Type | |
|---|---|---|
query | string | required |
kind | food | herb | any | optional, default any |
limit | integer | optional, 1–50, default 10 |
nutrient_ranking
Rank foods by a nutrient per 100 g — the "highest in iron" question. Each row also gives the typical serving heathy.org holds for the food and the amount in it; the order is per 100 g.
Foods eaten a few grams at a time — dried herbs, spices, tea leaves, and dried seaweed,
mushrooms and vegetables, 102 foods — are not ranked in results. They are reported
beside it in also_higher.small_amounts: a count of those holding more per 100 g than the
last row, and the first five. They are ranked in results when include_small_amounts is
true, and when the category asked for is mostly such foods (spice, tea, seaweed).
| Argument | Type | |
|---|---|---|
nutrient | string | required — slug or name |
limit | integer | optional, 1–100, default 20 |
category | string | optional — e.g. fruit, legume, millet |
min_sources | integer | optional, default 1 — smallest number of sources for a food to be ranked; foods below it are reported in also_higher.fewer_sources |
include_small_amounts | boolean | optional, default false — rank spices, tea and dried seaweed with every other food |
herb_monograph
An Ayurvedic herb monograph. Traditional use and modern research are kept strictly
separate — traditional records what the tradition claims, modern records what
research shows with an evidence tier. Every response carries a mandatory safety
block (cautions, pregnancy guidance, drug interactions).
| Argument | Type | |
|---|---|---|
herb | string | required — English, Sanskrit, Hindi or botanical name |
check_claim
For a raw whole food or an unfortified single-ingredient flour, which nutrition claims ("source of potassium", "high in vitamin C") the data supports under GB/EU rules (Reg. 1924/2006, per 100 g) and US rules (21 CFR 101.54, per RACC). It flags where the answer depends on which national table was used, listing the tables that do and don't support it.
| Argument | Type | |
|---|---|---|
food | string | required — slug or name |
nutrient | string | optional — check one nutrient only |
market | uk | us | both | optional, default both |
only_permitted | boolean | optional — return only supported claims |
preparation | raw | optional, and the only value accepted |
Scope. This is a screening tool, not permission to make a claim. It refuses cooked and manufactured products, whose composition depends on recipe and processing. A supported claim still needs an analysis of the producer's own produce: the law looks at the final product, not at a reference table. It lists the tables that do and do not support a claim; it does not choose one that qualifies.
search_recipes
Find original raw / no-cook vegan recipes on Heathy by free text, health goal,
category, ingredient, time or allergen. Per-serving nutrition is computed from the
same food rows lookup_food returns.
| Argument | Type | |
|---|---|---|
query | string | optional — free text, e.g. breakfast |
goal | string | optional — e.g. immune, bone, gut, heart |
category | string | optional — e.g. smoothies, salads, chutneys |
ingredient | string | optional — food slug or name that must appear |
max_time | integer | optional — total minutes |
exclude_allergens | string[] | optional — e.g. ["tree-nuts","sesame"] |
limit | integer | optional, 1–50, default 10 |
get_recipe
One recipe in full: ingredients with food slugs and gram weights, method, per-serving nutrition, allergens and carbon footprint.
| Argument | Type | |
|---|---|---|
recipe | string | required — slug or name |
health_topic
What to eat for a health goal — the nutrients involved, why each one, and which raw
recipes carry them. Accepts a goal ("gut health") or a named condition ("PCOS",
"anaemia"). Topics people reach for with a diagnosis carry a note on what the evidence
does and does not support, with its own evidence rating. Omit topic to list every
topic.
| Argument | Type | |
|---|---|---|
topic | string | optional — goal slug or title, or a condition name |
include_recipes | boolean | optional, default true |
Examples
| Call | Result |
|---|---|
search_foods("karela") | gourd-karela-raw |
lookup_food("guava", ["vitamin-c"]) | 222 mg · median of 13 tables · range 68–273 |
lookup_food("besan", ["protein"]) | chickpea-flour · 22.39 g · 3 tables |
lookup_food("lentils boiled") | refused — names lentil-whole-brown, held unprepared |
nutrient_ranking("iron", category: "legume", min_sources: 3) | brown lentil 8 mg · white soybean 7.91 mg · chickpea 5.4 mg |
nutrient_ranking("magnesium") | hulled hemp seed 591.5 mg (177.4 mg per 30 g) first · 37 foods eaten a few grams at a time counted beside the list, led by dried sea lettuce 3,200 mg |
check_claim("banana", nutrient: "potassium") | GB/EU "source of" supported at 335 mg — 16 of 21 tables support it and 5 don't |
health_topic("anaemia") | iron-deficiency-anaemia — nutrients, reasons, recipes |
herb_monograph("haldi") | turmeric — traditional, research, safety |
Response shape
Every tool returns the same envelope:
{
"data": { /* the payload */ },
"attribution": ["USDA FoodData Central …", "IFCT 2017, ICMR–NIN …"],
"source_url": "https://heathy.org/food/guava-flesh-only-raw/",
"disclaimer": "Informational only — not medical advice.",
"license": "https://heathy.org/terms/"
}
Each response includes attribution, disclaimer and source_url; herb responses
also include a safety block with cautions, pregnancy guidance and drug interactions.
The terms at https://heathy.org/terms/ ask that figures you republish carry the
attribution.
Data sources
Americas — USDA FoodData Central (Foundation Foods and SR Legacy) · Canadian Nutrient File · TACO (Brazil) · INCMNSZ (Mexico) · ARGENFOODS (Argentina)
Europe — UK CoFID · ANSES-Ciqual (France) · BLS (Germany) · Frida (Denmark) · Matvaretabellen (Norway) · Livsmedelsverket (Sweden) · Fineli (Finland) · NEVO (Netherlands) · CREA (Italy) · Swiss FCDB · NutriData (Estonia)
Asia — IFCT 2017 (ICMR–NIN, India) · MEXT (Japan) · China CFCT · Korea MFDS · Singapore SG FoodID · Pakistan FCT · Sri Lanka FCT
Africa and Oceania — FAO/INFOODS WAFCT (West Africa) · Kenya KFCT · Ethiopia EFCT · Australia AFCD (FSANZ) · Pacific Nutrient Database
Global — FAO/INFOODS BioFoodComp · open-licence (CC BY) research papers for foods no national table covers · Phenol-Explorer (polyphenol content, shown but not averaged)
Full credits and licence terms: https://heathy.org/sources/
Licence and terms
The files in this repository (documentation, manifest and Dockerfile) are released under the MIT licence. That licence covers this repository only. Composition figures come from the published tables above, each under its own terms, and are credited per response and at the link above. The reconciliation, recipes, herb monographs and original content are © Heathy — see https://heathy.org/terms/. Use of the hosted service at mcp.heathy.org is governed by those terms.
Not medical advice. The data is informational and does not diagnose or treat anything. A 100% raw-vegan diet does not supply adequate vitamin B12 or vitamin D.
Tools it offers (8)
What this server listed when Ahel dialed its public endpoint in Oct 2026, with no key and no account of yours. The names are the server’s own.
lookup_foodsearch_foodsnutrient_rankingherb_monographsearch_recipesget_recipehealth_topiccheck_claim
Signals
- Last commit
- Sep 2026
Advanced
- Delivery
- nutrition MCP server → your Ahel connector (mcp.ahel.ai) → your AI.
- Item type
- mcp-server
- Key
org-heathy-nutrition- Source
- github.com/chandata/heathy-mcp
- Hosted endpoint
https://mcp.heathy.org/mcp
github.com/chandata/heathy-mcp