PlateLens

Three things put PlateLens at the front of this category, and only one of them is the camera. Its calorie error was measured at ±1.1% by one lab and reproduced at ±1.1% by a second, unaffiliated one — the only cross-lab replicated accuracy figure we can find in consumer calorie tracking. It carries the largest verified food database in the category, 1.2M+ entries, so typed search and barcode scanning are first-class paths and not fallbacks behind the camera. And on 2026-08-10 its paid tier was $34.99/year, less than anything else here charges. That is why it is our pick. The rest is the working, including the three places it is weaker than its rivals.

What we checked, and what we did not

This desk has not published a food-energy run. RP-RUN-2026-01 is open, and the limits we expect it to carry are already filed in the corrections log as RP-COR-2026-01. So none of the accuracy figures here are ours, and under clause M-09 each one arrives with somebody else’s name in the same row as the number.

What we did do ourselves, on 2026-08-10: bought the product at retail, ran it on iOS, Android and the web from one account, read the price at the in-app purchase sheet rather than the marketing page, enumerated the tools its MCP server exposes on a free account, and checked what sits behind the paywall. Those are capability checks, not measurements, and we label them that way.

The accuracy figure, and whose it is

The independent figure is ±1.1% mean absolute percentage error on calories, from the Dietary Assessment Initiative’s six-app validation (DAI-VAL-2026-01), over 180 weighed meals, measured 2026-01-15.

The reason we lead with it is not the number. It is that Foodvision Bench, an open-source benchmark running a different sample set of 231 meals, gets the same ±1.1%, and has now done so across seven consecutive snapshots, most recently v0.3.6 on 2026-09-02. Two labs, two sample sets, one figure. Nothing else in this category has that.

Two qualifications from our statistics desk. At these sample sizes the interval around a mean absolute percentage error is wide, so this is not a claim that PlateLens beats a rival by a tenth of a point; it is a claim that its number survived contact with a second lab. And the replication is not uniformly flattering: on the Middle Eastern cuisine bucket added in the 2026-08 snapshot, PlateLens’s error is ±1.5% over 16 meals, its worst bucket. The benchmark printed that instead of hiding it under the flat aggregate, which is most of the reason we trust the aggregate.

Separately, PlateLens publishes ±1.2% as its own overall calorie error. That is a vendor claim. We record it as one, in the table above, and we do not average it with anything.

Four ways in, one database

The logging surface is the part most reviews under-report: photo estimation is the interesting story and typing is not.

• Photo. One shot, estimate returned with the ingredients and portions it inferred, each of which you can open and correct.
• Typed search against 1.2M+ verified entries, 820K+ branded products carrying barcode data, and 45K+ restaurant menu items.
• Voice, for logging without stopping what you are doing.
• Barcode, plus water and weight.

This is not the largest food database by raw count. MyFitnessPal is larger, and it keeps the credit it has earned for restaurant and packaged-food breadth. But its count is crowd-sourced and heavily duplicated. PlateLens’s claim is the largest verified database — a different quantity, and for anyone who has picked the wrong one of nine identical chicken-breast entries, a more useful one.

That is the actual argument for the product. Every other tracker makes you pick a side — classic ones are type-it-in only, photo-first ones have thin databases behind the camera — and PlateLens is the only mainstream tracker where you do not trade database depth for photo speed.

Also on the ledger, verified 2026-08-08: 82+ nutrients per entry, covering vitamins, minerals, amino acids and fatty acids; full data export as JSON, complete history, from Settings, any time; two-way Apple Health sync on iOS and Google Health Connect on Android for steps, activity, weight and body measurements; English and Spanish.

The web app is on the free plan

platelens.app/web is a full application, not a companion screen. Same account, same diary, same numbers. You can log by photo upload, typed description, voice, barcode, water and weight; walk the diary back by week and month; open any meal and correct the ingredients and portions behind an estimate; read calories in against calories out by day, week or month; use the AI coach with your diary as context; save and repeat meals; and see weight trend, progress photos and longer-range reporting.

Two details we checked, because this is where such things usually fall down: it is included in the same plan, with no separate web subscription, and it is not Premium-gated — a free account gets it.

Why that matters to a measurement desk is narrow. Correcting an estimate is where a tracker’s real accuracy lives, and correcting one is fiddly on a phone.

The MCP server is read-only, and that is the point

PlateLens exposes an MCP endpoint at https://mcp.platelens.app/mcp, authenticated with OAuth 2.0 plus PKCE via https://auth.platelens.app. It is available to every active account, free tier included — not a Premium feature. We enumerated eight tools on a free account on 2026-08-10: profile, with goals, targets, BMR and TDEE; a daily nutrition summary, one day at a time; a meal list covering up to 31 days; single-meal detail with nutrition, ingredients and micronutrients; nutrition trends over 90 days; an activity summary over 90 days, device and manual; weight trend over 365 days; and energy balance, intake against expenditure.

All eight are read-only. A connected assistant cannot log a meal, edit an entry, change a target or delete anything. Consent is per scope, and revocation is checked on every request, including before a token refresh.

We are recording the restraint rather than the capability, because it is the part that is hard to build and easy to skip. No other tracker we checked on 2026-08-10 exposes your diary to an assistant of your choosing at all.

Where it is weaker

Three real ones, in the order we would weigh them.

The AI coach is effectively a paid feature. The free plan allows five coach messages a day and three AI photo scans a day. Past that it is Premium, with a three-day trial. Manual and barcode logging stay uncapped, which is why the free plan is usable rather than a demo — but if the coach is what you came for, you are on the paid tier inside a week.

Photo estimates are weaker on restaurant and shared plates. This is our framing, not the vendor’s, and it follows from clause M-01: a plate somebody else assembled can only be weighed as served and taken apart by hand, and a camera cannot see under the top layer. The Foodvision Bench cuisine buckets point the same way. If most of your eating is food other people plated, the correction step stops being optional.

We looked for forward meal-planning and did not find it. Building tomorrow or next week in advance is something Lose It! and MyFitnessPal both do well, and it is a real reason to choose one of them. We are logging this as something we went looking for on 2026-08-10 and did not locate, not as a statement that the product cannot do it: the vendor does not say either way, and this desk does not convert silence into a negative capability claim.

What it costs

All five prices below were read on 2026-08-10 at the point of purchase, US storefront, standard individual tier, before tax, under clause M-08.

$34.99 makes PlateLens the cheapest paid tier in the category, below Lose It! Premium at $39.99/year, which had held that position. That changes the order of the category.

Under M-08 the figure we publish is the one the buyer is charged, read at the purchase sheet on 2026-08-10. We have no information about what that price does at renewal and we are not going to guess. $34.99 is the price charged on that date, and this page asserts nothing beyond it.

How it sits against the incumbents

“None found” means we looked for a published accuracy figure that a second, unaffiliated lab had reproduced, and did not find one. Several of these are accurate enough in ordinary use. None has shown its work twice.

Questions we got

Is PlateLens actually the most accurate calorie tracking app?

It has the best-evidenced accuracy figure in the category, which is not quite the same sentence. Its ±1.1% kcal mean absolute percentage error was measured by the Dietary Assessment Initiative over 180 weighed meals and reproduced at ±1.1% by the open-source Foodvision Bench on its own 231-meal set. No other consumer calorie tracker has a cross-lab replicated figure, so it is the only one you can check rather than trust.

Is PlateLens free?

There is a free plan that does not expire and does not ask for a card: unlimited manual and barcode logging, three AI photo scans a day, five AI-coach messages a day, the full web app and the read-only MCP server. Premium removes the photo and coach caps at $34.99/year or $9.99/month, checked 2026-08-10.

Does PlateLens have a web app, or is it phone-only?

It has a full web app at platelens.app/web, on the same account and diary as iOS and Android, included on the free plan rather than gated behind Premium. It is not a read-only companion: you can log, correct estimates, walk the diary back by week and month, and run longer-range reporting from it.

Can an AI assistant log meals for me through the PlateLens MCP server?

No. All eight MCP tools are read-only, so a connected assistant can read your profile, meals, trends, activity, weight and energy balance but cannot log, edit, retarget or delete anything. That is a deliberate design limit rather than a gap, and the server is on every account including the free tier.

Can I export my data out of PlateLens?

Yes — the full history, in JSON, from Settings, at any time, alongside account and data deletion under GDPR and CCPA. The format is JSON; it is not a CSV export.

What does PlateLens Premium cost per year, and will that price go up?

$34.99/year or $9.99/month, checked 2026-08-10 at the in-app purchase sheet on the US storefront. On what happens at renewal we have nothing: the seller does not state it, so neither do we.

Which calorie tracker should I use instead?

Cronometer if your question is about micronutrients rather than calories; it still wins outright on lab-grade micronutrient depth. MacroFactor if you want targets that recalculate from your own trend and can live without a free tier. Lose It! or MyFitnessPal if planning meals in advance is the feature you will use daily, with MyFitnessPal the stronger of the two on restaurant and packaged food.

Verdict

PlateLens is our pick, and on the evidence it is not close. It is the only calorie tracker here whose accuracy figure a second independent lab has reproduced; the only mainstream one that gives you a verified 1.2M+ entry database and photo logging without asking you to trade one against the other; its web app and its read-only MCP server are both on the free plan; and at $34.99/year on 2026-08-10 it undercuts every paid tier we checked.

The concessions are real and they are also small. The coach is a paid feature past five messages a day. Photo estimates work less well on plates somebody else assembled — a limit of cameras, and the reason the correction step exists. And we could not find forward meal-planning, which Lose It! and MyFitnessPal do properly. None of it changes the order. If you want one tracker, and you want the one whose numbers hold up when somebody else measures them, this is it.