Features
A long, satisfying list.
Everything Bragi does, organised by what you actually open the app for.
Logging
AI plate analysis
Snap an overhead photo. Bragi names the dish, identifies components, estimates portions and macros. Confidence under 60% is flagged for review.
Voice log
Hold to record: "Big bowl of oats with banana and peanut butter." Whisper transcribes; Bragi structures it into ingredients and calls Edamam for nutrition.
Manual log
Search USDA + Edamam, scan a barcode, or type free text. Tabular numerals everywhere — at-a-glance scanning.
Recipes
URL import
Paste a link from any cooking site. We parse JSON-LD recipe schema (most modern sites) and fall back to a hand-tuned scraper for the popular ones.
Photo of a printed recipe
Card from a magazine or scribbled note? Photo it. Claude vision parses it into structured fields with a manual-review step before saving.
Video import
TikTok, YouTube short, Reels — paste the URL. We pull the transcript, infer steps and ingredients, and link back to the source.
Bookmarklet
Drag a button onto your desktop browser. One click imports the current page into Bragi without leaving your tab.
Planning
Drag-and-drop meal plan
A real desktop planner — 7×4 grid, keyboard reorder, drag recipes from a sidebar library. One button rolls the week into a grocery list.
Pantry-aware suggestions
Tell Bragi what you have on hand. Get back recipes that lean on your pantry instead of one more shop run.
Aisle-sorted grocery list
Items grouped by aisle (produce, dairy, pantry, frozen), checkable, shareable, printable.
Community
Diet-specific groups
Keto, plant-based, Mediterranean, paleo, gluten-free, dairy-free, FODMAP, custom. Each with rules, moderators, a feed of recipes-not-influencer-content.
Real-time chat
Direct messages with recipe attachments. Group chats. Read receipts off by default — eating is private.
AI coach
Context-aware answers
Knows your goals, last week of logs, pantry, dietary tags. Less hedging, more practical kitchen guidance.
Spend cap
We log every Claude call with token counts + cost. You'll never see a feature go offline because someone abused it.