The clean keto kitchen

Every figure in this tool is computed from USDA FoodData Central, SR Legacy release 2018-04 — 99 of the 104 pantry ingredients map to a specific FoodData Central record. Nothing is hand-entered: energy, protein, fat and carbohydrate come straight from that record, and each recipe's macros are the sum of its ingredients divided by servings — including a dish's base recipes, which roll up the same way.

Net carbohydrate is Carbohydrate, by difference minus Fiber, total dietary. Quantities describe the raw ingredient — cooking drives off water and concentrates everything per gram, so a cooked portion of the same food reads higher. Millilitre amounts are converted using each food's own USDA portion — olive oil's 1 cup = 216 g gives 0.913 g/ml — rather than a density picked by hand.

Per-piece amounts use a USDA portion weight wherever one exists: one large egg = 50 g, one garlic clove = 3 g, one Hass avocado = 136 g of flesh, six medium oysters = 84 g. Four are assumptions, because USDA publishes no such portion — the 125 g mackerel tin (USDA lists only a 361 g can), and the parsley, cilantro and dill bunches, taken as 60, 40 and 20 g. Recipes use quarter-bunches of herbs, so those three shift a serving by a calorie or two at most. Every portion decision is recorded in the mapping file below.

The fat-quality bar is built from individual fatty-acid rows rather than estimated: marine omega-3 is EPA + DHA + DPA, plant omega-3 is ALA, omega-6 is linoleic + arachidonic + GLA, MCT is C8 + C10, and saturated is total saturated fat minus those MCTs. The bar shows the share of each class among the fatty acids USDA identifies, so the segments are proportions of identified fat, while the gram figures in the tooltip are the actual USDA amounts. A recipe earns the ω-3 rich badge at 0.5 g EPA+DHA+DPA per serving — twice the 250 mg daily adequate intake, and about the 500 mg often cited for cardiovascular benefit.

Three items have no SR Legacy equivalent — MCT oil, coconut aminos and bone broth — and use typical manufacturer label values instead. Two more are composites, noted because it matters: ghee takes its macros from the FoodData Central ghee record but its fat profile from butter, because the ghee record reports no monounsaturated fat and would otherwise read as pure saturated; and avocado-oil mayonnaise takes mayonnaise macros with avocado oil's fat profile. Cuts and varieties are named where they change the answer — the salmon here is farmed Atlantic at 208 kcal per 100 g, where wild is 142.

Two of the base recipes are special cases where summing the raw ingredients would mislead. Bone broth simmers bones you then throw away, so the bones carry no "as eaten" macros — they're on the shopping list only — and the broth's own numbers are finished-broth values. Sauerkraut likewise uses the fermented product's values, not raw cabbage, since fermentation changes it. Every other base is a straight sum of its ingredients, the same as any dish.

Ingredient names describe the record behind them rather than the nicest version of the food, which is why a few read oddly. The tinned mackerel is jack mackerel, leaner than the Atlantic mackerel most tins actually contain; the smoked salmon is chinook; and the sauerkraut figures come from canned kraut — still buy it raw and unpasteurised for the live cultures, just don't expect the macros to differ. Where a name is broader than its record — olive oil, sea salt, chilli flakes — the difference doesn't move the numbers.

You do not have to take any of this on trust: the full ingredient-to-FoodData-Central mapping, with the record ID, the exact USDA description and the computed values for every one of the 104 items, is published as ingredient-sources.json, alongside the script that generated it. That script is self-contained — it downloads the USDA release itself and rebuilds the whole table — so anyone can reproduce these numbers from scratch.

These are reference values for generic foods, not measurements of what you actually bought. Real ingredients vary with season, breed, feed, ripeness and brand. Treat the numbers as a well-sourced compass, not a lab result.