Skip to content

Next.js 15/Tailwind 4 · Spec Harness AI layer

Vibe-code the site. Don't rebuild it when it grows.

AI gets you a landing page in an afternoon. Then it grows, the vibe-coded site buckles under real traffic and real SEO, and you rebuild it properly a month later. This is the foundation for what comes after: a clean, AI-indexed Next.js 15 codebase, SEO and AEO ready, with the whole idea-to-deployed workflow built in. Any engineer can pick it up.

Read the docs

One-time €99 · own the code · lifetime updates · deploy unlimited sites


See it running

Fifty seconds, the real product.

No mockups. This is the boilerplate you download, recorded as it runs: landing page, dashboard, docs and the component kit.

01/The problem

Vibe coding is a black box.

It gets you a v1 in a weekend. Then it stops scaling. You cannot see why the AI made the choices it made, so every new prompt is a fresh roll of the dice on a codebase you do not fully understand. Three things break, in order.

  • 01

    AI hallucinates on a blank repo

    Drop Claude Code or Cursor into a fresh template and it guesses your conventions: invented file paths, wrong data flow, APIs that do not exist. You spend the afternoon undoing it.

  • 02

    Generic templates give UI, not context

    v0, Lovable, and starter kits hand you components. None of them ship the rules, memory, and architecture an AI agent needs to extend the code reliably.

  • 03

    Tangled architecture compounds

    Every shortcut you take, the AI copies. Without enforced structure, the codebase drifts a little more with each prompt until nothing is safe to touch.

The fix is not a better prompt. It is opening the box: a codebase your AI can see through, plus the workflow that keeps it on spec. Built for founders, indie hackers, and agencies who ship websites with Claude Code or Cursor and are tired of cleaning up after the blank repo.

02/The answer

A codebase your AI can see through.

AI Web Launcher ships the code and the workflow on top of it. Three things turn a black box into something an agent builds on reliably.

01

Structure

A production Next.js 15 codebase with a fixed shape: typed config, env-gated services, feature folders. The AI fills a known layout instead of inventing one.

02

Context

The Spec Harness memory bank and CLAUDE.md brief load on every session, so the agent reads your conventions before it writes a line. No more cold starts.

03

Guardrails

Enforceable rules, a spec with checkable criteria, and one atomic build task at a time. Drift gets caught before it compounds into a rewrite.

Here is how that compares to a generic template or starting from scratch.

03/Honest comparison

How it stacks up.

AI Web LauncherGeneric templateFrom scratch
Modern UI + sections out of the box
Blog + docs engine included
Partial
SEO / JSON-LD / sitemap pipeline
Partial
Lead capture + email + Stripe wired
Partial
Zero-config (builds with no .env)
Partial
Scalability (feature-first structure)
Partial
Performance (Core Web Vitals ready)
Partial
Clean code architecture
Partial
Idea → deployed workflow (9 steps)
AI knows your codebase (Spec Harness context layer, no hallucinating)
Recommended skills + MCP setup
Battle-tested in production
Partial
Time to first deploy
An afternoon
A day or two
Weeks

04/What's included

The codebase, fully wired. Then the workflow on top of it.

A complete, brandable Next.js foundation that builds green with zero environment variables. The workflow (next section) is what turns it into your site without the AI guessing.

Next.js 15 + TypeScript

App Router, React 19 Server Components, strict TypeScript. A modern foundation with sensible defaults.

Tailwind 4 + shadcn/ui theming

CSS-first Tailwind v4 with shadcn/ui OKLCH tokens. Re-skin the whole site from one token block. Light and dark themes included.

Responsive landing + sections

Hero, social proof, problem, features, comparison, pricing, FAQ: composable, mobile-first marketing sections.

Blog engine

A typed blog with categories, authors, and OG images. Ships with a sample post you can repurpose as a changelog.

Custom /docs engine

A flat .tsx docs system with callouts, copy-ready code blocks, steps, and a clean sidebar; the docs you're reading run on it.

SEO + JSON-LD + sitemap

getSEOTags helpers, structured data, OG/Twitter image routes, automatic sitemap, robots, and IndexNow pings.

Supabase lead capture

Auth, database, and a fail-soft /api/waitlist lead route, env-gated, so it never crashes when unconfigured.

Resend email

Transactional email wired in and disabled cleanly until you add a key. Same fail-soft pattern across the stack.

Stripe-ready checkout

A one-time checkout route that falls back to email capture when Stripe isn't configured. This site sells itself with it.

Configurable admin dashboard

A /dashboard shell with a collapsible sidebar, search, notifications, a user menu, six analytics charts and a settings page. Public by default, or hide it behind a Supabase auth token.

Spec Harness ai_rules/ system

The memory bank, enforceable rules, generators, and decision records that give AI agents real context on your code.

Claude Code rules + skills + MCP

A CLAUDE.md brief, a .cursorrules file, recommended skills, and MCP config so your agent works on day one.

One-click Vercel deploy

Push, import, deploy. Zero build config. The postbuild step generates your sitemap automatically.

0
workflow steps, idea to deployed
0
services wired end to end
0
documentation pages in the box

The nine artifacts the workflow writes

concept-brief.mdPRD.mdcopy.mdglobals.css tokens.memory/00-70src/features/.mcp.jsonshipped featureslive URL
concept-brief.mdPRD.mdcopy.mdglobals.css tokens.memory/00-70src/features/.mcp.jsonshipped featureslive URL
concept-brief.mdPRD.mdcopy.mdglobals.css tokens.memory/00-70src/features/.mcp.jsonshipped featureslive URL
concept-brief.mdPRD.mdcopy.mdglobals.css tokens.memory/00-70src/features/.mcp.jsonshipped featureslive URL

What's inside the dashboard

A real admin shell, in the box.

Most templates stop at a marketing page. This one ships a working /dashboard — nine composable components and six charts, all driven by config and themed with the same tokens as the rest of the app.

Collapsible sidebar

Config-driven nav with grouped sections, badges, an active state that tracks the route, and an icon-rail collapse. You pass data, not markup.

Command palette

A grouped ⌘K / Ctrl-K search over your pages and docs. Keyboard-first, wired to the router, open from anywhere.

Top bar

Sticky header with the search trigger, a notifications menu with an unread count, the user menu, and a slot for per-page actions.

Six analytics charts

Area, bar, line, stacked-bar, donut and radial — each takes a single data prop, so the same set powers a real usage dashboard.

Create modal

A dialog with inline validation and a success toast. Copy the pattern for any create-or-edit flow in your app.

Settings form

Profile picture upload, name, username, email and bio, with client-side validation. Wire it to your API and ship.

Light and dark

Every surface reads from the Colorway C tokens — one accent, hairline structure, off-black and off-white. Flip the mode, nothing breaks.

Auth-ready

Public by default so you can demo it, or hide the whole route behind the Supabase auth token in one line.

9 components·6 charts·⌘K palette·light + dark

Read the dashboard docs

05/How it works

The workflow is the product. The code is the easy part.

AI made writing code cheap. It did not make deciding what to build cheap, or make a model that invents your file paths reliable. So the value moved up a layer, to the chain that keeps the code on spec. Three phases, nine steps, idea to deployed.

01/DefineYou + an LLM
01Idea to briefconcept-brief.md
02Brief to PRDPRD.md
03PRD to copycopy.md
04Design accentglobals.css
02/InitializeThe agent
05Memory bank.memory/00-70
06Architecturesrc/features/
07MCP + skills.mcp.json
03/BuildThe agent
08SDD build loopshipped features
09Deploy + re-runlive URL

Every step writes a file the next one reads. That ordered handoff is what you're buying. See the full workflow in the docs.

06/The differentiator

Built for you and your AI.

The workflow runs on four things that ship in the box. Every other template stops at UI; these are the rules, skills, MCP setup, and memory bank that give Claude Code and Cursor real context instead of a guess.

Spec Harness memory bank

The ai_rules/ directory: enforceable coding rules, generators, decision records, and a context map, plus the always-loaded CLAUDE.md brief. Your agent reads it before writing a line.

Read the Spec Harness docs

Claude Code rules

Domain-specific rules for Next.js, SEO, design, and performance, written as enforceable instructions, not vibes. The AI follows your conventions, not generic defaults.

See the rules

Recommended skills

A curated set of Claude Code skills that pair with this stack (frontend design, SEO audits, humanizing copy, drafting blog posts) and how to add your own.

Browse the skills

MCP setup

Pre-picked Model Context Protocol servers (Supabase, Context7, Playwright) so your agent can query your schema, pull version-accurate docs, and verify pages in a real browser.

Configure MCP

07/Pricing

One price. Everything included. Forever.

No tiers, no subscription, no per-seat math. Buy once, then re-run the workflow for every site you build, your own products or paid client work.

Lifetime deal

AI Web Launcher Lifetime

The full boilerplate, the AI-indexing layer, and every future update. One payment, deploy unlimited sites.

Founding access just opened · first 200 at €99, then €199

19989one-time
  • Complete Next.js 15 + Tailwind 4 + shadcn/ui codebase
  • Blog engine + custom .tsx /docs + SEO/JSON-LD pipeline
  • Supabase leads, Resend email, Stripe-ready checkout
  • Spec Harness ai_rules/ memory bank + Claude Code rules
  • Recommended skills + MCP setup, ready to go
  • Lifetime updates, every future release included
  • Deploy unlimited sites, no per-project license

Code

  • Next.js 15 App Router, React 19 Server Components, strict TypeScript
  • Tailwind 4 + shadcn/ui OKLCH tokens, light and dark themes
  • Responsive landing, blog engine, and a custom .tsx /docs engine
  • SEO helpers, JSON-LD, OG/Twitter routes, sitemap, robots, IndexNow
  • Env-gated Supabase, Resend, and Stripe, so it builds with no .env
  • One-click Vercel deploy with automatic sitemap on postbuild

Workflow

  • The 9-step idea-to-deployed chain, each step writing a file the next reads
  • Concept brief and PRD generators that make done machine-checkable
  • Copy generator that writes on-voice marketing strings
  • Design-system generator: two knobs re-skin the whole site
  • The SDD build loop: specify, plan, tasks, implement, one task at a time
  • Re-runnable for every site you build, your own or paid client work

AI layer

  • Spec Harness ai_rules/ memory bank: rules, generators, and decision records
  • CLAUDE.md brief and .cursorrules so Claude Code and Cursor have context
  • A curated set of recommended Claude Code skills for this stack
  • MCP setup: Supabase, Context7, Playwright, and 21st.dev Magic

Support

  • Lifetime updates, every future release included
  • Personal onboarding to get your first site shipped
  • Deploy unlimited sites, no per-project license

Instant access · lifetime updates · personal onboarding

The Ship-It Guarantee

If you don't have a site running within 7 days, email me and I'll get you unstuck. Still not for you? 50% back within 7 days, and you keep the repo.

Shipping mobile too? Your purchase gets you 30% off AI Mobile Launcher.
Using Wire AI in your product? Take 30% off with code WIREAI30.

07/FAQ

Questions, answered.

What exactly do I get for €99?
Two things. First, the full AI Web Launcher codebase: a production Next.js 15 app (Tailwind 4, shadcn/ui, blog engine, custom /docs, SEO/JSON-LD/sitemap/IndexNow, Supabase lead capture, Resend email, Stripe-ready checkout). Second, and the real point, the orchestrated workflow that ships it: nine ordered steps from a one-paragraph idea to a deployed site, each writing a file the next step reads. That includes the Spec Harness ai_rules/ memory bank, the prompts and generators, the Claude Code rules, recommended skills, and MCP setup. One payment, lifetime updates, re-run it for unlimited sites.
Why does the workflow matter more than the code?
AI made writing code cheap. It did not make deciding what to build cheap, and it did not make a model that invents your file paths reliable. The value moved up a layer: from typing code to running the chain that keeps the code correct, on spec, and consistent. The workflow forces the decisions (brief), makes 'done' machine-checkable (PRD), keeps copy and design on-brand (generators), stops the agent re-learning the repo every session (memory bank), and builds against the spec one atomic task at a time. The full chain is documented step by step at /docs/workflow.
What's the stack?
Next.js 15 (App Router, React 19), Tailwind CSS 4 with shadcn/ui OKLCH design tokens, Geist fonts, a custom .tsx /docs engine, Supabase for auth and leads, Resend for email, Stripe for checkout, and next-sitemap. Every integration is environment-gated, so the project builds and runs with no .env file at all.
Do I need to know AI or prompt engineering to use it?
No. You can clone it, edit one config file, and deploy like any template. The AI-indexing layer is a bonus: if you use Claude Code or Cursor, the included rules, skills, and memory bank make the agent reliable on the codebase. If you don't, it's still a clean, modern Next.js starter.
Can I use Cursor instead of Claude Code?
Yes. The boilerplate ships both a .cursorrules file and a CLAUDE.md memory bank, plus the shared ai_rules/ pack. Cursor, Claude Code, Windsurf, and most agentic editors read this context and generate code that matches the architecture.
What is Spec Harness?
Spec Harness is the AI-context layer in the ai_rules/ directory: enforceable coding rules, repeatable generators, decision records, and a context map, anchored by the root CLAUDE.md. It's what stops AI agents from hallucinating file paths, APIs, and conventions on your codebase.
What is the best Next.js boilerplate for Claude Code and Cursor?
One that ships agent context, not just compatibility. AI Web Launcher is a Next.js boilerplate AI can actually build on: the Spec Harness ai_rules/ memory bank, CLAUDE.md, .cursorrules, and generators mean Claude Code and Cursor read the architecture before writing code, instead of inventing file paths and conventions on a repo they've never seen.
Is there an AI-ready alternative to ShipFast?
That's the gap this product exists for. ShipFast is the most established Next.js boilerplate and works with AI editors, but it doesn't ship an AI-indexing layer. We keep an honest side-by-side at /vs/shipfast, including the cases where ShipFast is the better buy (JavaScript, Pages Router, MongoDB, the bigger community).
Do I get updates?
Yes — lifetime updates are included. Every future release of the boilerplate (new integrations, framework bumps, improved rules) is yours at no extra cost.
What's the license? Can I use it on client work?
One purchase lets you deploy unlimited sites, including paid client projects and your own products. You can't resell the boilerplate itself as a competing template.
Do you offer refunds?
You keep the full GitHub repository from day one, so there are no full refunds. But if it's not what you expected, email us within the first 7 days and you get 50% back, no questions asked.

Stop fighting your boilerplate. Ship the website.

Clone it, brand it in one config file, and let your AI build on a codebase that actually has context. One payment, every future update, unlimited sites.

Read the docs