Open Animal Nutrition

One open way to check a dog or cat diet against the published nutrient requirements: the same result, in the same words, for every person and every AI that uses it.

• Open the checker
• Read the Implementation Guide
• Help govern it

Why it exists

Diets for companion animals, especially dogs and cats, are formulated by animal caretakers, veterinarians, animal nutritionists, veterinary nutritionists, and companies, and, more and more often, suggested by AI assistants, whose recommendations are not always accurate. Whether a diet meets a requirement comes down to comparing its nutrient concentrations with a standard's figures: the NRC's, AAFCO's, or FEDIAF's. That comparison is easy to get wrong, by hand or by a model working from memory, and the standards give minimums and maximums, not names for where a concentration lies between them.

Open Animal Nutrition writes the comparison down as one open computation, so that a diet checked by anyone, with any tool that follows it, gets the same result.

How it works

1. Choose one requirement: one of 21 from the 2006 NRC Nutrient Requirements of Dogs and Cats, the 2026 AAFCO Official Publication and the FEDIAF Nutritional Guidelines of September 2025. Future standards can be added as published or needed, including from other jurisdictions.
2. Enter the diet's nutrient values and, where the requirement uses them, facts about the animal, such as its body weight and life stage.
3. For each nutrient, the checker reports where the value lies among the 14 bands of the Dynamic Decile System, from "Very Severely Restricted (<25% of Min)" to "Highest (90% of Range or Greater)", and a proposed verdict: deficient, excess or none.

Every maximum is labelled by kind, and only a value above a maximum that a standards body set for that animal and use is called excess. Where a standard gives no maximum, the range high is set or computed by a disclosed rule, most often four times the minimum, for consistency, not as a toxicity limit.

Every result first says what it is not: not a safety or compliance assessment, and not veterinary advice. Only the nutrients entered are checked.

Why shared words matter

The standards state their figures in different ways (per unit of energy, of dry matter or of body weight) and set different minimums and maximums. Two people comparing the same diet with the same standard can still describe the result differently: "low", "marginal", "borderline".

The Dynamic Decile System puts every nutrient concentration on one scale, four bands below the minimum and ten deciles of the span from the minimum to the range high (the last also takes values above it), and gives each band one name. "Slightly Restricted (75 to <100% of Min)" or "Near Highest (80 to <90% of Range)" names the same position against the selected standard's limits in every report, for every standard and to every reader; it is a position, not a judgment of safety. That shared vocabulary lets people compare diets, standards and recommendations like for like.

It matters even more for AI. A language model may not have a standard's figures in its training data, may recall them imprecisely, and can answer the same question differently each time. Through the checker's MCP connector (the standard way for AI assistants to call outside tools), an assistant sends a diet's values to the same computation a person uses and gets back the same result every time: each nutrient's band, its verdict, and the conditions the verdict depends on, in the system's own words. The checker grounds the assistant in a computation instead of its memory; it does not replace the judgment of the veterinarian or nutritionist using it.

Help govern it

The rules are a proposal. They become a standard only when independent experts have checked, challenged, and ratified them. The Dynamic Decile System table was reviewed by the steward's advisory board and, informally, by more than 60 board-certified veterinary nutritionists and nutrition residents. Those reviews were of versions up to September 2025; neither group has reviewed the changes made since, including new inputs, several changed limits and the checker's own rules.

Who we are inviting: board-certified veterinary nutritionists (ACVIM Nutrition, ECVCN), members of the AAVN, nutrition residents, animal nutrition scientists, and other groups of veterinary nutrition experts.

What the body decides first:

1. the common basis: the Dynamic Decile System's bands;
2. the rule for maximums where a standard sets none;
3. the rules by which a nutrient's concentration and limits give its verdict;

and then how the project is hosted and funded, and whether to accept transfer of the domain, the GitHub organization and the hosting account from the provisional steward. The guide also lists the points still open and the conflicts found in applying the standards' figures, for the body to resolve.

What it takes: no programming. The first questions are nutritional; the later ones (hosting, funding, transfer) are decisions, not technical work. The computation and its rules are written out in plain language, and every change from the template is listed with its reason. AI-powered assistance can help with most technical issues that may involve GitHub, code, hosting, and documentation.

To take part, or to suggest someone, please email openanimalnutrition@dvmconsulting.com. Please pass this page on to colleagues who should be part of it.

Where to find things

Who runs it now

The provisional steward is Davis Veterinary Medical (DVM) Consulting, Inc., a California Certified B Corporation; Sean Delaney, BS, DVM, MS, DACVIM (Nutrition), acts for it. It sells nutritional supplements and offers free diet formulation software that implements this standard; Sean Delaney is the lead co-editor of Applied Veterinary Clinical Nutrition, 2nd edition, and an author of two other works the checker cites (disclosure of interest). The computation references no product, ingredient, supplement or vendor. DVM Consulting pays for the domain and hosting until a governance body decides how the project is hosted and funded, and intends to transfer stewardship to a neutral, multi-stakeholder body. Everything is CC0: anyone may use, change or republish it.

Privacy

This page and the Implementation Guide are served by GitHub Pages. GitHub logs the IP address of every visitor for security, whether or not they are signed in to GitHub, under the GitHub Privacy Statement; the steward receives no record of visits. These pages set no cookies and use no analytics. This page runs no scripts. The Implementation Guide's pages are generated by HL7's FHIR publishing tools: they run scripts served from this site, which remember display choices (such as an open tab) in your browser's storage, on your device, and a few of them load a help icon from hl7.org.

The checker, where you enter a diet, is a separate service at checker.openanimalnutrition.org, on Amazon Lightsail. In normal use it keeps no record of what you enter or of IP addresses; the one exception is an unexpected failure, when its log, kept for 3 days, may hold part of the request. Its About page says exactly what it keeps. Through an AI assistant, what you type in the assistant you use is covered by that provider's terms, not by this site.

Why GitHub. The project's code, data, issues and reviews are already on GitHub. Serving these pages from there means each version of the guide is built in public, by a workflow anyone can inspect, and publishing a new version takes a change merged on GitHub rather than running a server, which the volunteer experts of a governance body should not need to do. These pages also stay up if the checker has to be taken offline, and they move with the GitHub organization when stewardship is transferred. The cost is a record of visitors' IP addresses, held by GitHub. We accepted it for pages that only publish information; the checker, which receives what you enter, keeps its stricter promise.