FAQ

Foodpile is an MCP-first tracker. Instead of tapping through a food database, you tell your AI assistant (Claude, ChatGPT, Mistral) what you ate, did, or weigh, and it logs it over Foodpile’s MCP server. The website is a datastore plus a dashboard that shows your day, trends and forecasts.

No. Foodpile is a datastore, not an AI service — there are no server-side AI calls and no models running on Foodpile. The intelligence lives in the assistant you already use; your data is never sold.

Yes. The dashboard cards have an in-app add (the header +) and per-row edit/delete for caffeine, water, weight and measurements, so quick corrections don’t need a chat. The assistant is still best for anything in natural language.

It models caffeine in your body as first-order exponential decay: each dose adds its milligrams at the time you had it, and the total falls by half every half-life (5 hours by default, adjustable). The curve is the sum of all the day’s doses decaying over time — solid up to now, dashed for the projection.

It’s the sleep threshold — the level below which caffeine is unlikely to disturb sleep. The card tells you the clock time your level is projected to drop below it, so you can judge a late coffee against bedtime.

A "what if I had a coffee now?" preview: they drop a hypothetical dose at the current time so you can see how the rest of the day (and bedtime) would look. It is a preview only — nothing is logged and nothing is sent to the server. You can hide the experiment in settings.

It says "this day was maintenance" — you neither gained nor lost ground on it. In the calorie-balance forecast the day is counted as a day at exactly your maintenance TDEE: whatever was (or wasn’t) logged is discarded, and the day contributes no deficit of its own. Use it for a day whose food logging was incomplete or off — a half-logged day (breakfast only) would otherwise look like a huge deficit and make the forecast over-optimistic.

Yes, slightly — and that is deliberate. Counting the day at maintenance dilutes your average pace by the share of days that were neutral: 15 logged days at a 500 kcal/day deficit plus 5 neutral days averages out to 375 kcal/day. Simply deleting those days would instead model a future in which off-days never happen again, which quietly makes every goal date too optimistic for anyone with a recurring rest or holiday day. The card names the neutral days in its basis line so the number always adds up.

No. The scale-trend forecast reads your weigh-ins, and a weigh-in is empirical — the scale said what it said. Marking a day neutral is a statement about food-logging quality, so it changes only the calorie-balance forecast and the trend averages. That separation is also why the two estimates are worth having side by side.

Mark it anyway. An unlogged day is normally invisible to the forecast — it is simply absent, so a week away would leave your pace looking untouched. Marking it makes the day count as maintenance, which is usually much closer to the truth for a holiday or a day off tracking. On the charts such a day shows a flat mark on the baseline, since there is no bar to fade.

Three ways: the toggle in the today card on the dashboard, tapping a day’s bar on the trends page, or asking your assistant ("mark yesterday as incomplete"). A marked day shows faded on the trend bars, with a flat mark on the baseline.

No — a fast is a real, intended deficit, so fasting days stay as real data and are never auto-marked. Neutral marking is a manual signal that a day’s data is unreliable, not that you ate little on purpose.

They are two independent views of the same goal. One reads the scale trend (a least-squares line through your weigh-ins). The other reads the calorie balance (your mean intake over recent logged days vs. your maintenance TDEE). When they agree you can trust the date; when they disagree that itself is informative — e.g. the scale hasn’t caught up with a new deficit yet.

The calorie-balance estimate averages your logged days, so one under-logged day pulls it. Mark such days neutral (see above) and it settles — they then count as maintenance days instead of as enormous deficits. The forecast is descriptive, not a promise — it just extends your current pace.

Waist/hips/chest/arm/thigh are the fat-sensitive companion to the scale: during recomposition the scale can sit still for weeks while a falling waist shows the real progress. Each site is its own line on the measurements chart.

A static goal is a fixed daily number. A dynamic goal is a base plus the calories you burn from logged activity — so a workout raises the day’s target ("eat back" your training). Pick whichever matches how you like to plan.

Fibre is tracked as its own macro dimension and totalled like protein/carbs/fat. Because it is a subset of carbohydrate, it isn’t part of the calorie-balance split — it just gets its own bar (and a broad 30 g/day fallback when you haven’t set a target).

Yes — set your unit system and energy unit in your profile. Values are stored canonically (kg, cm, kcal) and shown in your unit everywhere, including what the assistant reports back.

No — conversions keep a decimal, so an inch value converted to cm and back lands on the same number. The same care applies to weight (kg/lb) and energy (kcal/kJ).

Six charts over a period you choose — calories, macros, caffeine, activity, water and weight — all driven by one selector (week / month / 3 months / year). A year is shown as weekly averages so it stays legible.

Tap a day’s bar to open that day on the dashboard or mark it neutral; toggle the macro series (protein/carbs/fat) in and out of the stack; and read the per-chart summary line for the period’s averages.