AI Nutrition Planner for Trainers

AI Prompts to Build This

Copy these build prompts into Claude, Cursor, or your AI coding tool. Create a free account to unlock the full research behind them.

1. 1. Project Setup

   Build the weekend MVP of the "AI Nutrition Planner for Trainers": a trainer enters a client's goal and macro targets and gets a week of meals whose macros are checked against USDA data and corrected before the trainer sees them. It supports a trainer and is not medical nutrition advice.
   
   Stack: Next.js (App Router, TypeScript), Tailwind, Supabase (Postgres, Row Level Security, Auth with email magic link for trainers), the Anthropic API to write the plan and fix a day, the USDA FoodData Central API to check macros. Clients have no login: the trainer prints or shares the plan page. Deploy on Vercel.
   
   Tables (Row Level Security on, every row is scoped to its trainer):
   - trainers(user_id, brand_name, brand_primary_hex)
   - clients(id, trainer_id, name, goal, target_calories, target_protein_g, target_carbs_g, target_fat_g, allergies jsonb, cuisines jsonb, budget_weekly) goal is cut, bulk, recomp or maintain
   - plans(id, client_id, week_start, target_macros jsonb, actual_macros jsonb, status, llm_model, created_at)
   - meals(id, plan_id, day, meal_type, name, ingredients jsonb, macros jsonb, prep_minutes, recipe_text)
   
   Screens: /login, /clients, /clients/new, /clients/[id]/plan (generate and review), /plan/[planId] (a printable client page in the trainer's colors).
   Env vars (names only): NEXT_PUBLIC_SUPABASE_URL, NEXT_PUBLIC_SUPABASE_ANON_KEY, SUPABASE_SERVICE_ROLE_KEY (server only), ANTHROPIC_API_KEY, USDA_API_KEY.
   
   Do not build: billing, plans or trainer tiers, a client portal or adherence tracking, grocery shortcuts, Slack alerts, a PDF library.
   Done when: npm run dev starts, you can sign in, and the four tables exist with Row Level Security on.

2. 2. Core Feature

   Build the one feature that proves the Planner: macros the trainer can trust, shown against the target, with allergies never broken.
   1. The client form takes a goal, target calories and macros, allergies, cuisines and a weekly budget. Refuse a calorie target below a minimum kept in one config file, and tell the trainer to refer that client to a registered dietitian.
   2. Ask the Anthropic API for a 7-day plan as strict JSON { days: [{ day, meals: [{ type, name, ingredients: [{ name, grams }], recipe_text, prep_minutes }] }] } with the targets, the allergies, the cuisines and the budget in the prompt.
   3. Check allergies in code before anything else: no ingredient may match a client allergen or its common synonyms. Reject the plan and regenerate once if one does.
   4. Validate macros: for every ingredient, search USDA FoodData Central in the Foundation and SR Legacy data, take the best match, scale its per-100-gram values to the grams and sum each day. An ingredient with no match is flagged unknown and left out of the totals.
   5. If a day is more than 5 percent from any target, ask the model to change one named meal, for example "Day 3 has 180 g protein against 150 g; replace one lunch with something lower in protein at the same calories". Replace only that meal and re-check. Stop after 3 rounds and keep the best result.
   6. Save target_macros and actual_macros on the plan and show both, with the meals that were flagged unknown.
   7. The printable plan page shows the week in the trainer's name and primary color.
   Rules: say on the plan "Prepared by your trainer. Not medical nutrition advice. Tell your trainer about any medical condition." Never show a macro as exact when an ingredient is unknown.
   Empty state: with no clients, show the form and a sample profile.
   Done when: a sample client gets a plan with no allergen, a day that is 10 percent off is corrected within 3 rounds or flagged, the plan shows target and actual side by side and an unknown ingredient is called out.

3. 3. Landing Page

   Build a one-page landing site for the AI Nutrition Planner, meal plans for personal trainers' clients.
   Hero: "A week of meals that hits your client's macros." Sub: "Enter the goal and the targets. Get a plan checked against USDA food data, with allergies respected, ready to send in your name." One button: Join the waitlist.
   Sections: a sample plan page next to its target and actual macros, how it works in three steps (client profile, generate and check, send), and an FAQ on how macros are checked, what the plan is not (medical nutrition advice) and which tools you can send it with.
   Waitlist: store the email in a waitlist table in Supabase. No other service.
   Style: Geist, an off-white background, ink type, one fresh green accent.
   Done when: the page renders on a phone and a submitted email appears in the waitlist table.

4. 4. Branding Package

   Use a design or image tool for this one. A coding agent cannot draw a logo.
   Brand for the AI Nutrition Planner: pick a short working name, then design a wordmark and an icon that suggest a plate divided into three macro segments.
   Colors: off-white, ink and one fresh green. Type: Geist, with tabular numbers for macros.
   Deliverables: wordmark, icon, a macro bar style for protein, carbs and fat that differs by pattern as well as color, a printable plan header, and one launch graphic.
   Done when: each deliverable is saved in one folder and the macro bars are distinguishable without color.