Native iOS · SwiftUI AI · on-device capture

Pantry AI.

Recipes from what you already have

Point your camera at the fridge, or just say what you've got. Pantry AI reads your ingredients with LLM vision and generates a real recipe — macros, steps, and a shopping list for whatever's missing.

Swift SwiftUI SwiftData LLM Vision iCloud / CloudKit Sign in with Apple
Pantry AI — app home screen
01 How it works

Three taps from fridge to dinner.

No manual entry, no hunting through recipe sites. Capture what you have, let the model do the thinking, then cook and save.

1 CAPTURE

Scan or speak

Snap a photo of your ingredients or dictate them by voice. Vision + on-device capture turn the real world into a structured ingredient list.

2 GENERATE

LLM does the cooking logic

An LLM with vision reads the ingredients and returns a complete recipe — steps, portions, and macros — tuned to what you actually have on hand.

3 COOK & SAVE

Scale, shop, sync

Scale servings, auto-build a shopping list for what's missing, favorite it, and sync across devices with iCloud.

02 Inside the app

A full, native iOS app — not a web wrapper.

Vision recognition

Scan ingredients straight from the camera or photo library — no typing required.

Voice input

Just say what's in the kitchen and let the app build the list for you.

Smart shopping lists

Auto-generate a list of the ingredients you're missing for any recipe.

Scaling & substitutions

Resize any recipe to your serving count and surface ingredient alternatives.

iCloud sync

Recipes and favorites follow you across iPhone and iPad via CloudKit.

Sign in with Apple

Private, secure account management built the way Apple intends.

03 Engineering

How it's built.

App

LanguageSwift
UISwiftUI
ArchitectureMVVM
PersistenceSwiftData
TargetiOS 17+

Intelligence & sync

RecipesLLM (text)
IngredientsLLM vision
CaptureCamera · Speech
SynciCloud / CloudKit
AuthSign in with Apple

Want a closer look?

Pantry AI is one of several apps I've shipped solo. Happy to walk through the build, the vision pipeline, or the SwiftData model live.