MeetingMood AI

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 "MeetingMood AI": a calendar-connected tracker where you rate last week's meetings and see which ones drain you.
   
   Stack: Next.js, TypeScript, Tailwind, shadcn/ui, Supabase (Postgres, Row Level Security, Auth with Google sign-in). Deploy on Vercel.
   Calendar: Google Calendar only, read-only scope, requested in the same Google sign-in.
   
   Tables (Row Level Security on, so each user reads only their own rows):
   - calendar_connections(id, user_id, provider, refresh_token_encrypted, connected_at)
   - events(id, user_id, external_id, start_at, end_at, attendees jsonb, recurring_series_id, title_hash)
   - reactions(id, event_id, user_id, sentiment, created_at) where sentiment is one of like, meh, drain
   - weekly_reports(id, user_id, week_start, summary_markdown, top_drain_series_id)
   
   Screens: /login, /review, /dashboard.
   Env vars (names only): NEXT_PUBLIC_SUPABASE_URL, NEXT_PUBLIC_SUPABASE_ANON_KEY, SUPABASE_SERVICE_ROLE_KEY (server only), GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET.
   
   Do not build: billing or plans, teams or manager views, Microsoft 365, an admin area, email digests.
   Done when: npm run dev starts, you can sign in with Google, and the four tables exist with Row Level Security on.

2. 2. Core Feature

   Build the one feature that proves MeetingMood AI: the weekly swipe review and the patterns it reveals.
   1. On first sign-in, pull the last 4 weeks of events from Google Calendar (expand recurring events into single events) and insert them into events. Repeat each Sunday night with a scheduled job that uses the service role key.
   2. /review shows one card per meeting from the last week, newest first. Three buttons do what a swipe does: like, meh, drain. The keys L, M and D work too. Each answer writes one row to reactions.
   3. /dashboard runs three plain SQL group-bys over reactions joined to events: average sentiment by weekday, by recurring_series_id, and by number of meetings that day. Show the worst weekday and the worst recurring series as two cards.
   4. Write one template sentence about the top drain series into weekly_reports.summary_markdown. No AI call.
   Output shape: { by_weekday, by_series, top_drain_series } as JSON from one route handler.
   Empty state: with fewer than 10 reactions, show "Rate 10 meetings to see your first pattern" and the review button.
   Done when: you can sign in, rate 10 past meetings in /review, and /dashboard names your worst weekday and your worst recurring series.

3. 3. Landing Page

   Build a one-page landing site for MeetingMood AI, a mood tracker for your meeting calendar.
   Hero: "Find out which meetings are draining you. With data." One button: Join the waitlist.
   Sections: the problem (a week that feels off with no single cause), the swipe review (a short animated demo of like, meh and drain), a sample weekly report with names hidden, a privacy section (read-only calendar access, and your reactions are yours alone), and a short FAQ.
   Waitlist: store the email in a waitlist table in Supabase. No other service.
   Palette: ink near-black, warm paper #fcfaf7, lavender #b6a4f6 for the drain line.
   Voice: calm, data-savvy, slightly dry.
   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.
   Wordmark for MeetingMood AI with a small swipe-arc curving under the "M", readable at 18px.
   Colors: ink #0F1115, paper #FCFAF7, lavender #B6A4F6 for drain charts, sage #9BC5A8 for energy.
   Typography: Geist Sans for the interface, Geist Mono for chart labels.
   Deliverables: favicon, a "your first week starts Sunday" empty state, a set of swipe-card illustrations (cardstock paper feel, soft shadows, no faces), and one launch graphic.
   Done when: each deliverable is saved in one folder and the wordmark is readable at 18px.