Build your own Cal AI with Claude Code

Build my own take on Cal AI: an AI calorie counter that works from photos. Give it an original name, logo, colours and copy, and do not use Cal AI's name, branding or assets.

Version one should have:
- Snap a photo of a meal
- AI estimates calories and macros
- Edit the estimate before saving
- Daily totals against goals
- History and streaks

The hard part: Accuracy. Use a vision model, show its confidence and make corrections one tap away.

How to build it:
1. Read this folder first. If it is empty, set up Next.js (App Router, TypeScript, Tailwind) with Postgres through Prisma or Supabase, and Supabase Auth or Auth.js for sign-in. Write a short PLAN.md with the data model, screens, third-party services and what you are leaving out of version one.
2. Build the core loop end to end before anything else: snap a photo of a meal, working with real data saved to the database.
3. Add the rest of the features one at a time, and seed realistic demo data so every screen looks alive.
4. Keep API keys in environment variables, and put any paid service behind a small interface with a local stub, so the app runs on a fresh machine without accounts.
5. Run it, click through every screen on desktop and mobile widths, fix what breaks, and finish with a summary of what works, what is stubbed and how to deploy it.

Paste it into Claude Code in an empty folder. It plans first, so read PLAN.md and steer before it starts building.