A production-ready Next.js starter for AI coding tools. Every screen already works on mock data; your agent wires the real services one seam at a time.
$225 once · works with Claude Code, Cursor, Codex, Gemini CLI and Copilot
The skills are plain Markdown. Any agent that can read a file and run a command can wire this kit.
The same prompt, run twice.
Once against an empty repository, once against the kit. Staged, but fair: this is the shape each starting point gives your agent.
Structure, colors and copy invented mid-prompt. The next hour goes to fixing the drift.
One plan price from your mock data, ready for Stripe.
The screen already existed; tokens and hierarchy came from DESIGN.md. The only real work left is wiring checkout.
The screens, on the table.
Marketing site, auth, dashboard, billing, search, files, settings, blog, docs and legal. Everything renders before you add a single API key.
Sign-in is already done.
Every auth screen ships finished on mock data. Login, register, reset and verification render on day one; your agent connects a provider to one marked seam instead of inventing forms, sessions and edge cases.
- Password and magic link
Sign-in, reset and verification screens that already work end to end.
- Social sign-in
Google, GitHub and Apple behind the same provider seam.
- Roles in the model
Owners, admins and members in the mock data, ready for your permission rules.
- Safe defaults written down
The wiring skill carries the env var names and security defaults, so nothing is invented under pressure.
Take payments without building payments.
Pricing, checkout, portal and subscriptions are finished screens. Stripe wires in first, and the seam is provider-shaped, so you can add Polar or Lemon Squeezy later by remapping the same seam.
- Hosted checkout
A secure hosted checkout, so there are no card forms for you to build.
- Plans, invoices and portal
Subscription management and billing history, already shaped and clickable.
- Swap providers later
The same flows, with a new provider mapped onto the existing seam.
Search that answers, with receipts.
Semantic search is a finished product surface, not a bolted-on chat widget. It renders ranked, cited results on mock data from the first run; connect your embeddings model and content when you are ready.
- Ranked by meaning
Results ordered by how well they answer the question, even when the words differ.
- Bring your own model
OpenAI, Anthropic or Gemini behind one key, swapped without touching the UI.
- Sources and citations
Every answer names the files it used, so people can check the work.
- Relevance you can see
Scores surface in the UI, so tuning retrieval is a loop you can watch.
The rest of the kit, indexed.
Each capability ships finished on mock data, with one plain-Markdown skill that walks your agent through wiring it.
26 skills ship in the repo, from contact forms to security audits. Every one is a Markdown file in agent-skills/, greppable like everything else.
Pick one of 300+ designs. Your kit arrives wearing it.
Every design on duply.ai ships as a DESIGN.md, and your license covers the one you pick: the same kit, delivered in that exact visual language. Because one token file drives all 29 screens, rebranding it to your own product later is an edit, not a rewrite.
- One file, not a theme system
DESIGN.md holds the whole visual language as plain role tokens.
- Nothing hardcoded
Components read tokens only, so there is no stray hex to chase down.
- Dark mode included
DESIGN-DARK.md carries the second mode; the same build wires both.
skip the plumbing.
The bill you are not going to pay.
Ten line items, estimated at ordinary contractor rates. Disagree with the rate? Slide it. The kit costs $225 either way.
Pay once. It is your repo after that.
No subscription, no seat math, no locked editor. One license covers unlimited projects, including the ones you ship for clients.
- Lifetime license, unlimited projects
- Client work included
- Full source, no black boxes
- Updates as versioned zips, no re-buy
hosted checkout via Polar · card, Apple Pay, Cash App
Auth, billing, dashboard, files, settings and routing: 29 finished screens.
Semantic search with cited sources, bring your own model.
Blog, docs, SEO, llms.txt, legal, contact, newsletter and analytics.
DESIGN.md tokens, dark mode, and the exact design you pick from the library.
26 agent skills, 41 marked seams, and convention files for Claude, Codex and Gemini.
Everything you saw on this page ships in the kit. This site is the kit.
Asked before buying.
The short version: real code, your repo, works with your agent, no subscription.
The full Next.js source for the kit: every screen built and clickable on mock data, the integration seams clearly marked, and a matching agent skill for each one. It arrives as a zip, restyled in the design you picked.
No. It is real, editable code in your own repo. There is no proprietary editor or runtime to escape from, so you keep working the way you already do.
Yes. The screens already work on mock data, and each integration has a plain-language skill that walks an AI agent through wiring it. You describe what you want; the agent follows the skill.
From-scratch prompting drifts and leaves half-built screens. Here the structure, design system and seams are already decided, so the agent fills in known gaps instead of inventing the whole app.
Any agent that can read files and run commands. The skills are plain Markdown, so Claude Code, Codex, Cursor, Gemini CLI and Copilot all read them the same way.
Run the dev server and click through the working app. Then say "set up my kit": the built-in onboarding interviews you and rebrands the kit around your product.
Yes. It is your repo to edit, fork and ship. The license is one-time, and updates arrive as versioned zips.
Yes. One lifetime license covers unlimited projects, including work you build and ship for clients.
Stop negotiating with an empty folder. Pick a design, get the kit, and let your agent start from finished.
$225 once · unlimited projects · full source