BUILD WITH AI.BREAK NOTHING.

The Next.js AI SaaS starter built for agent-driven development.

SET UP
BUILD
VERIFY
SHIP
Built withNext.jsSupabaseStripeMulti-model AIClaude · OpenAI · Gemini
[STANDARDIZE · VERIFY · OWN]

ADOPT AI AGENTS
WITHOUT LOSING CONTROL

AI agents at their full potential.

  1. 01

    STANDARDIZE

    Contract files (CLAUDE.md, AGENTS.md) are architecture policy as code. Every developer's agent reads the same rules, so ten people produce one codebase, not ten.

  2. 02

    VERIFY

    The verification suite is plain npm scripts. Run it locally, in review, or on every pull request. Agent-written changes get checked before they merge.

  3. 03

    OWN

    Full source in your private repo, perpetual license, no runtime dependency on us. If Whirlwind vanished tomorrow, your product keeps working unchanged.

npm run doctor -- --deep
  supabase   ✓ reachable · migrations applied
  stripe     ✓ key valid · test mode
  ai         ✗ "claude-3-5-sonnet" not found (likely retired).
             Set ANTHROPIC_MODEL to a current model.
npm run auditconfig, styles, security
npm run doctor -- --deeplive services + model IDs
npm testsignup, checkout, credits
33static checks
26end-to-end tests
3providers pinged
6migrations
How teams use WhirlwindRead the verification docs →
[WHAT'S INCLUDED]

SIX CORE
SYSTEMS

Everything a production SaaS needs, already wired.

[01]

AGENT-DRIVEN SETUP

Your agent walks you through setup and explains each step. Contract files teach the agent your architecture, so new code lands where it belongs.

[CAPABILITIES]
CLAUDE CODE · CURSOR · OPENCLAW
REVIEWED SCRIPTS, NOT PROMPTS
SCRIPTS WRITE, THE AGENT EXPLAINS
NO AGENT REQUIRED
[WORKS WITH ANY AI AGENT]

Setup is a conversation, not a guide. Claude Code, Cursor, or OpenClaw, whichever you already use, walks you through setup and explains each step, so you spend the first hour deciding what to build instead of reading docs. One command confirms the result.

  • Reviewed scripts do the writing and migrations. The agent explains and verifies.
  • Keys are typed into the setup script in your terminal, not the chat, and land in a gitignored .env.local.
~10 MINto a running app
Also wiredBlog·Docs·SEO + sitemap·Email (lifecycle + Mailchimp)·Support chat (Crisp)·Analytics (Plausible)·Admin dashboard·Rate limiting
[PRICING]

ONE PAYMENT,
PERPETUAL LICENSE

Buy it once. No subscription, no lock-in.

LAUNCH OFFER
$100 OFF[LIMITED TIME]

BUILD

$250
$150
USD ONE TIME
SAVE $100
  • Complete Whirlwind codebase
  • Agent-driven setup (works with any AI agent)
  • Verification suite: static audit, live doctor, e2e tests
  • Multi-model AI client (Claude, GPT, Gemini)
  • 3 working AI features: chat, writer, image studio
  • Supabase auth + database, Stripe payments
  • Refund handling + production deploy guide
  • Lifetime updates
  • 1 commercial project
7-DAY REFUND
INSTANT DELIVERY
[RECOMMENDED]

ACCELERATE

$300
$200
USD ONE TIME
SAVE $100
  • Everything in BUILD, plus:
  • Unlimited commercial projects
  • Priority email support
  • Early access to new features
7-DAY REFUND
INSTANT DELIVERY

TEAM

$600
$500
USD ONE TIME
SAVE $100
  • Everything in ACCELERATE, plus:
  • Up to 5 developer seats (reassignable)
  • Unlimited commercial projects for your org
  • One shared architecture: contract files + CI checks
  • Invoice / PO purchasing available
7-DAY REFUND
INSTANT DELIVERY

More than 5 seats, or licensing for an agency?

Larger orgs and client work priced directly · invoice / PO purchasing available on TEAM and above

Whirlwind for TeamsContact us
[STYLE PRESETS]

YOUR PROJECT
TRANSFORMED

Same codebase. Different product. Drag the divider to compare.

Pick an example product for the right-hand side.

Any color, any preset, light or dark. All of it is one config file.

Style
Color
Theme
Palette
GalleryPricingAPIGenerate one
for brand teams
Brand assets, on demand

Describe the shot and Palette renders it in your own colors, type and crop ratios, so everything your team ships already sits on brief.

Generate oneSee the gallery
12,430
assets this week
6
brand kits
4s
per render
✦launch banner, product on a plinth, brand palette
rendering…
5 of 6 approved
How Palette works
Brief
Lock the palette, type and ratios once.
Generate
Variations that stay inside the brief.
Approve
Ship to the asset library in a click.
Pick a plan
Starter
200 renders / mo
$29
Studio
1,500 renders / mo
$99
Agency
Unlimited
$299
© PalettePrivacy · Terms
Whirlwind
FeaturesPricingDocsGet started
[ your product ]
Your AI product, ready to ship

Auth, payments, credits and a multi-model AI client, wired together and audited after every change.

Get startedFeatures
6
migrations
26
e2e tests
1
command to verify
Everything wired
Auth
Magic links, RLS and middleware, already wired.
Payments
Checkout-first Stripe flow with refunds handled.
AI Client
Claude, GPT and Gemini behind one interface.
Simple pricing
Starter
$29
100 credits
· 1 project
· Email support
Pro
$79
500 credits
· 5 projects
· Priority support
Team
$199
Unlimited
· Unlimited
· Shared billing
© WhirlwindPrivacy · Terms

Set in config/theme.js · Outfit · Inter · Palette is an illustration of what the template becomes, not a live deployment.

[SUPPORT]

FREQUENTLY ASKED
QUESTIONS

Licensing, teams, and how verification works.

  • A complete Next.js 14 SaaS template:

    • ▸Supabase auth (passwordless magic links, row-level security)
    • ▸Stripe payments: one-time and subscription, with refund handling
    • ▸Multi-model AI client (Anthropic, OpenAI, Google)
    • ▸3 working AI features out of the box: streaming chat, content writer, image studio
    • ▸Tailwind + DaisyUI with style presets and any primary colour, so your app won't look like a template
    • ▸Agent-driven setup: an instruction file for any AI agent (Claude Code, Cursor, OpenClaw) that knows the codebase and walks you through the setup scripts. Or set it up yourself.
    • ▸Verification suite: npm run audit (static checks), npm run doctor (live service & model checks) and npm test (Playwright e2e), so every AI-written change is verifiable before it ships

    You also get lifetime updates and access to the private GitHub repo.

  • Two mechanisms, before and after the code is written. Before: contract files (CLAUDE.md, AGENTS.md, an OpenClaw skill) teach your agent the architecture (where migrations, routes and components go, and which invariants it must not touch), so features land on-concept instead of drifting into a parallel codebase.

    After: the verification suite re-checks everything, including code the agent just wrote. npm run audit runs its static checks over config, security and drift; npm run doctor -- --deep pings live services and confirms your configured AI models (Claude, GPT, Gemini) still exist; npm test runs Playwright against signup, checkout and credits.

    The loop is: build with your agent, run the checks, ship. If an AI session broke webhook signing or an RLS policy, you find out in seconds, not from a customer.

  • You should use Claude Code, on top of Whirlwind. Agents are excellent at UI and features, and statistically shaky at exactly the things that hurt: webhook signature edge cases, RLS policies that actually deny, refund revocation, credit race conditions. Those mistakes are subtle: the app works in the demo and fails with real money.

    Whirlwind ships those parts already built and reviewed, plus the contract files (CLAUDE.md, AGENTS.md, a Cursor rule, an OpenClaw skill) that teach your agent the rules of the codebase, plus averification suite (audit · doctor · test) that proves the setup, and everything built after it, is still correct. Your agent builds the product that's actually yours, and the plumbing is already done and stays verifiable.

  • JavaScript, deliberately. Plain JS keeps agent-driven edits simple and diff-reviewable, and every API route validates input at runtime (which TypeScript alone wouldn't give you anyway). The codebase is small and cleanly separated, so converting incrementally to TS is straightforward if your team requires it.

  • Two emails arrive after payment:
    1. A sign-in link to your dashboard
    2. A GitHub invite to the private Whirlwind repo

    Accept the GitHub invite, clone the repo, and the dashboard has a 5-minute quickstart guide.

  • You can run it before you configure anything. npm run dev serves your landing page, blog and docs with no keys set. Configuration turns features on, it isn’t a prerequisite for seeing the product.

    Two commands do the work: npm run setup prompts you for each key, writes .env.local, and runs the database migrations, and npm run doctor verifies it worked. You run them in your own terminal. An AI agent can walk you through where each key lives and verify the result.

    The agent gathers your product decisions by chatting, and the reviewed scripts do the writing and migrating. Keys are typed at the script’s prompts, not into the chat. Works with Claude Code, Cursor, OpenClaw, or no agent at all.

    Once setup’s done, the same agent can scaffold new features (migration, API route, and component) in the right places, wired correctly.

  • Basic familiarity helps but isn’t required. Setup walks you through signup and key creation one service at a time, and npm run setup writes the .env file and runs your migrations for you, and you just paste each key when asked. If you can do that, you can ship Whirlwind.

  • Whirlwind ships two things: deterministic setup scripts (npm run setup, doctor, deploy) and a short instruction file for each agent: CLAUDE.md for Claude Code, AGENTS.md for Cursor and Codex, and a skill for OpenClaw.

    Point your agent at the repo. The agent reads its file, walks you through the scripts, and verifies the result. The agent handles the conversation. The scripts do the actual work, so switching agents doesn’t change what happens under the hood.

    So no, you don’t need OpenClaw specifically. Use whichever agent you already have, or skip the agent entirely and run npm run setup yourself.

  • No. You paste them at the setup script’s prompts in your own terminal. The script writes a gitignored .env.local with chmod 600 and never prints a value back. Use Stripe and Supabase test keys during setup, and npm run doctor verifies the result. Everything runs on your machine.

  • 7-day no-questions-asked refund. Email hello@whirlwindai.xyz with your order info. We’ll refund you in full and revoke your repo access, and that’s it.

  • The BUILD plan licenses Whirlwind for one (1) commercial project. Building a second product with it requires upgrading to ACCELERATE (unlimited projects) or buying a second BUILD license. Both plans grant a perpetual license to the source code under the Whirlwind Commercial License so you can build as much as you want with what you have, but you can’t resell or redistribute the template itself.

  • Yes, and the verification suite was practically built for it. The contract files act as architecture policy every developer’s agent follows, and the verification suite is plain npm scripts, so you can run npm run audit and npm test on every pull request in your CI. Agent-written code can’t merge without passing.

    Licensing: the TEAM plan ($500, one-time) covers up to 5 developer seats (reassignable as your team changes) with unlimited commercial projects for your org and invoice/PO purchasing available. Individual licenses stay per-purchaser (repo access can’t be shared). More than 5 seats, or agency use across clients? Contact us or see Whirlwind for Teams.

  • Yes. The GitHub invite gives you ongoing access to the repo, including all future commits. We push updates as Anthropic/Stripe/Supabase APIs evolve, plus new features and fixes.

Still have questions?

CONTACT SUPPORT