Food Facts in ChatGPT
Look up food by barcode. The MCP server is https://food.openkrill.app/mcp. No key. The steps below are only for ChatGPT. A call to lookup_barcode was checked against that server on 2026-10-01.
What this server answers
Look up a packaged food by its barcode, search for products, compare them side by side and check a product's listed allergens.
Ask "what is in the product with barcode 3017620422003?", "find oat milks sold in France", "which of these three barcodes has the least sugar?" or "does this product contain milk or nuts?".
Food Facts reads Open Food Facts, a free database that people around the world add to by scanning and photographing labels.
A product answer shows the name, brand, size, ingredients, the allergens and traces the entry lists, the Nutri-Score and NOVA group when known, and nutrition per 100 g (energy, fat, saturated fat, carbohydrates, sugars, fibre, protein, salt and sodium).
Because the data is crowd-sourced, it can be incomplete, wrong or out of date, and recipes change. Every answer says how complete the entry looks and reminds you that the physical label wins, especially for allergies.
The allergen check reads the entry's listed allergen tags and its ingredients text. It is a data lookup, not a safety guarantee, and not dietary or medical advice.
Coverage is strongest in France and Europe and thinner in the US and Asia. For recalls of food products, use the Recalls plugin.
Data © Open Food Facts contributors, available under the Open Database Licence (ODbL) v1.0. It stores nothing about you: answers are cached for a few hours, and the only lasting record is a daily count of calls per tool.
What it can do
- Looks up a product by EAN or UPC barcode after checking the barcode's check digit
- Shows ingredients, listed allergens and traces, Nutri-Score and NOVA group
- Shows nutrition per 100 g: energy, fat, saturated fat, carbohydrates, sugars, fibre, protein, salt, sodium
- Searches products by name or brand, optionally limited to a country
- Compares up to five products by sugars, salt or energy per 100 g
- Checks a product's listed allergens against the ones you want to avoid
- Says how complete the entry is and reminds you to check the physical label
Tools
lookup_barcode, Look up a barcode. Use this when the user gives a product barcode (EAN or UPC) and wants to know what is in it, such as "what is in 3017620422003?". Returns the product name, brands, quantity, ingredients text, the allergens and traces the entry lists, Nutri-Score, NOVA group, nutrition per 100 g, a data-quality note and the Open Food Facts link. The barcode check digit is verified first. The data is crowd-sourced and may be incomplete or wrong: check the physical label. Not dietary or medical advice.search_products, Search products. Use this when the user wants to find food products by name or brand, such as "find oat milks" or "Nutella jars in Germany". Pass the words to search for and optionally a country (English name or two-letter code) and a limit of up to 10. Returns each product's barcode, name, brands and Nutri-Score; use lookup_barcode for the full entry.compare_products, Compare products. Use this when the user wants to compare packaged foods, such as "which of these barcodes has the least sugar?". Pass one to five barcodes and optionally sugars (default), salt or kcal. Returns a ranking from lowest to highest per 100 g with each product's other nutrition, Nutri-Score and NOVA group, and lists barcodes that were invalid, not found or lacked data. Crowd-sourced data: check the physical labels.check_allergens, Check allergens. Use this when the user asks whether a product contains certain allergens, such as "does 5449000000996 contain milk or nuts?". Pass the barcode and the allergens to avoid. Returns which are listed as contained, which as possible traces, which are not listed, and whether the entry has no allergen data at all. It reads the entry's allergen tags and ingredients text, which are crowd-sourced and can be wrong: it is a data lookup, not a safety guarantee. The physical label wins.
Add Food Facts in ChatGPT
ChatGPT does not read a local config file. A custom MCP server is added in the product, and only after developer mode is on. These steps were checked against OpenAI's plugin connect guide and the ChatGPT MCP guide on 2026-10-01. Developer mode can be hidden by a workspace policy, so a Team or Enterprise admin may have to allow it first.
- Open ChatGPT settings, then Security and login, and turn on Developer mode.
- Open ChatGPT plugins and select the plus button.
- Enter a name such as "Food Facts" and a one-line description of what you want it to answer.
- Under Connection, choose the public endpoint and paste
https://food.openkrill.app/mcp. Include the/mcppath. Do not add a tunnel id: this server is already public. - Create the connection and review the tools ChatGPT lists. You should see
lookup_barcode,search_products,compare_products,check_allergens. The feedback tools on the same server are optional and not required for the lookup. - Start a new conversation, add the connection from the tools menu, and ask: What is in the product with barcode 3017620422003?
No API key and no OAuth client id are required. The server answers without a login. Calls are limited per network and per day, and a limit comes back as a tool error rather than a partial answer. If the tool list is stale after a server change, open the connection and select Refresh, then start a new conversation.
The ChatGPT desktop app, Codex CLI and IDE extension share a different config. In the desktop app, open Settings, then MCP servers, choose Add server, select Streamable HTTP, and paste the same URL. In ~/.codex/config.toml the equivalent is a table named mcp_servers.food with url = "https://food.openkrill.app/mcp". ChatGPT on the web does not read that file. A published directory listing is also a different path: this page is the developer-mode connection, not a claim that Food Facts is already in the ChatGPT directory.
Treat the server as untrusted until you have read a few answers. A remote MCP server sees the arguments the model sends. Food Facts is read-only and documents what it sends onward on the privacy page.
A call checked on 2026-10-01
3017620422003 is the Nutella barcode in the plugin's starter prompts. The physical label still wins. The request below was posted to https://food.openkrill.app/mcp as tools/call. HTTP 200. Source: the Food Facts server, read 2026-10-01.
{
"method": "tools/call",
"params": {
"name": "lookup_barcode",
"arguments": {
"barcode": "3017620422003"
}
}
}- status. ok
- name. Nutella
- notice. Open Food Facts is crowd-sourced, so entries can be incomplete, wrong or out of date, and manufacturers change recipes. The physical label always wins: check it before you rely on this, especially for an allergy. Not die
- url. https://world.openfoodfacts.org/product/3017620422003
Repeat the call yourself if you need a newer reading. Cached answers expire. A rate limit is not a result: wait and try again. Nothing in the call is a ranking, a filing, or advice.