Storefront for Custom AI Agents

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 "AgentBay": a storefront where a creator publishes an AI agent and a buyer tries it through a metered proxy.
   
   Stack: Next.js (App Router), TypeScript, Tailwind, Supabase (Postgres, Row Level Security, Auth with email magic link). Deploy on Vercel.
   
   Tables (Row Level Security on: creators write their own agents, the public reads published listings):
   - creators(id, user_id, display_name)
   - agents(id, creator_id, name, description, endpoint_url, auth_header_name, auth_header_value_encrypted, pricing_model, price_cents, status) where pricing_model is one of flat, metered, tiered
   - listings(id, agent_id, slug, hero_copy, demo_input, category)
   - usage_events(id, agent_id, buyer_id, tokens_in, tokens_out, latency_ms, cost_cents, created_at)
   
   Screens: /login, /publish, / (discovery feed).
   Env vars (names only): NEXT_PUBLIC_SUPABASE_URL, NEXT_PUBLIC_SUPABASE_ANON_KEY, SUPABASE_SERVICE_ROLE_KEY (server only), ENCRYPTION_KEY.
   
   Do not build: payments or payouts, a second login, workspaces or agencies, a separate edge proxy, PII redaction.
   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 AgentBay: publish an agent, then call it through a metered proxy.
   1. /publish is a 4-step form: connect the endpoint and auth header, define the input and output JSON, pick a pricing model and price, confirm. It writes agents and listings, encrypts the auth header at rest, and makes one test call that must return 200.
   2. POST /api/proxy/[slug] checks the buyer is signed in, decrypts the creator's auth header, forwards the request, streams the response back, and writes a usage_events row with tokens, latency and cost.
   3. Limit each buyer to 20 calls per minute and return 429 above that.
   4. / lists published agents grouped by use case (writing, research, sales, ops, support, code). /a/[slug] shows the description, a live demo that calls the proxy with the sample input, the price and the creator's name.
   Rules: never return the stored auth header to any client.
   Empty state: with no published agents, show a button that starts /publish.
   Done when: a creator publishes an agent, a second account calls it from its product page, and the call appears in usage_events.

3. 3. Landing Page

   Build a one-page landing site for AgentBay, a storefront for custom AI agents.
   Hero: "Buy and run live AI agents. No glue code." One button: Join the waitlist.
   Sections: the problem (builders are stuck between GitHub stars and a real business), how it works in four steps (connect, price, publish, sell), a live demo mock of an agent product page, and a short FAQ on metering and trust.
   Waitlist: store the email in a waitlist table in Supabase. No other service.
   Style: Geist and Tailwind, white background, black accents, generous whitespace, no purple gradients.
   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 AgentBay: a wordmark and a small icon suggesting a doorway or storefront for agents.
   Colors: a near-black, one accent (warm teal or muted electric blue) and two neutrals. Type: Geist for the interface and headings, IBM Plex Mono for endpoint URLs and code.
   Voice rules: blunt and honest toward developers, never "revolutionary", always show prices in USD with cents.
   Deliverables: wordmark, icon, 8 product icons (connect, price, publish, demo, sell, meter, payout, analytics), and a one-page brand sheet.
   Done when: each deliverable is saved in one folder and the icons read at 24 px.