The AI application starter kit
for the Claude Code era.
Clean code enforced. Self-hosted ready. Production-grade from day one.
Stop burning hours on auth, vector DBs, and LLM glue. Clone, point Claude Code or Cursor at it, and ship your actual AI product.
Building AI applications shouldn't hurt this much.
Setup eats your day
Every new AI project burns 2–3 days on auth, DB, LLM clients, vector stores, and file upload. Tokens burned on scaffolding instead of business logic.
AI creates spaghetti
Claude Code and Cursor create 2000-line monster files with no structure and no tests. You refactor. They regress. You refactor again.
Demo breaks in production
Your demo works. Then production reveals twenty things that break — no rate limits, no cost tracking, no self-hosted path, no observability.
Clone, extend, ship. In that order.
A production-grade starter kit for full-stack devs building AI applications with Claude Code or Cursor. Clean code enforced. Self-hosted ready. Four LLM providers wired up.
AI-assistant native
CLAUDE.md, AGENTS.md, and .cursorrules files baked in. Your AI reads the conventions and extends business logic instead of re-scaffolding every new project.
Clean code enforced
ESLint rules, pre-commit hooks, and file-size limits make messy code literally fail to commit. 2000-line monster files? Not here.
Self-hosted ready
Four LLM providers: OpenAI, Anthropic, Gemini, and any OpenAI-compatible endpoint. Works with Ollama, vLLM, TogetherAI, Groq out of the box.
Production from day one
Rate limiting, cost tracking, error monitoring, structured logging, environment validation. The twenty things your demo would break on — already handled.
Modular & composable
Clean separation of platform primitives and feature modules. Add a new feature in minutes by following the pattern. Easy to extend, easy to own.
Docs your AI reads
Ready-to-paste Claude Code prompts in /docs/prompts/. Ask 'add a document summary feature' and it's created correctly — because every existing module is structured the same way.
From zero to production AI app in three steps.
Clone the repo
Get the full stack: Next.js, Drizzle, Better-auth, LLM client, vector DB, chat UI, production essentials. Everything wired and tested.
Point Claude Code at it
Your AI assistant reads CLAUDE.md and extends the codebase using the mandatory feature-module pattern. No re-scaffolding. No spaghetti.
Ship your AI app
Clean code. Self-hosted ready. Rate limits, cost tracking, error monitoring — all baked in. From idea to production in one codebase.
Everything you need. Nothing you don't.
The Stack
- Next.js 15 (App Router) + TypeScript
- Drizzle ORM + MySQL
- Better-auth (email + OAuth)
- S3 upload + PDF extraction
- Upstash Redis rate limiting
AI Infrastructure
- 4 LLM providers (OpenAI, Anthropic, Gemini, OpenAI-compatible)
- Abstract LLMClient interface
- Streaming responses
- Vector stores: pgvector + Pinecone
- Complete RAG pipeline
- Token + cost tracking
Quality Layer
- ESLint + strict TypeScript
- Pre-commit hooks (Husky + lint-staged)
- Max file size enforcement
- Clean-code rules documented for AI
- /docs/prompts ready-to-paste
Production
- Rate limiting + cost budgets
- Observability (Sentry, PostHog, Datadog)
- Environment validation
- Error monitoring hooks
- Production-ready auth
UI
- shadcn/ui base components
- 3 curated themes
- Streaming chat UI with citations
- File upload with progress
Plug-and-play
- Dockerfile + docker-compose
- Makefile with common commands
- Storybook support
- CI workflows
Built for full-stack devs shipping real AI products.
✅ Perfect for you if:
- You build AI apps with Claude Code or Cursor as your primary tool
- You're tired of re-scaffolding every new project
- You value shipping fast AND shipping clean
- You've fought the 'demo works, prod breaks' battle
- You need self-hosted LLM support (enterprise, compliance, cost)
- You want production-ready patterns from minute one
❌ Probably not for you if:
- You're learning your first framework
- You want a generic SaaS template (try ShipFast)
- You don't use AI assistants for coding
- You want a no-code visual builder
Why not just use...?
| UltraKit | ShipFast | Makerkit | OSS kits | |
|---|---|---|---|---|
| Production essentials | ✅ | ✅ | ✅ | ❌ |
| Lifetime updates | ✅ | ✅ | ✅ | ❌ |
| Commercial support | ✅ | ✅ | ✅ | ❌ |
| Pre-commit quality gates | ✅ | partial | partial | ❌ |
| AI-focused | ✅ | ❌ | partial | varies |
| 4 LLM providers built-in | ✅ | ❌ | partial | varies |
| RAG pipeline included | ✅ | ❌ | partial | varies |
| AI-assistant optimized | ✅ | ❌ | ❌ | ❌ |
| Clean code enforced | ✅ | ❌ | ❌ | ❌ |
| Self-hosted LLM adapter | ✅ | ❌ | ❌ | varies |
Simple, one-time pricing.
No subscriptions. Lifetime v1.x updates. Own what you pay for.
🎉 EARLY BIRD — First 30 buyers get 50% off. Lock in forever below.
Indie
Non-commercial, personal learning
- Full boilerplate
- All 4 LLM providers
- Clean code guardrails
- Discord access
- Lifetime v1.x updates
Commercial
1-developer commercial license
- Everything in Indie
- Commercial use
- Priority email support (24h)
- Private GitHub discussions
- 30-min Q&A call
Team
5-developer commercial license
- Everything in Commercial
- 5-dev license
- Priority feature requests
- Team onboarding (1h)
- Lifetime v2 updates
Questions you might be asking.
Built by someone shipping AI apps every week.
Built by an experienced full-stack engineer who ships AI applications for a living. 10+ years of experience spanning consumer-scale systems and enterprise AI work — including building the exact kind of boilerplate you're looking at right now, over and over again, on real production projects.
I built this starter kit because I got tired of re-scaffolding the same LLM client, vector store, auth, and file upload every time I started a new AI project. Every feature came from real pain, not theory.
This ships under a brand, not a face — my day job requires it. But the code is open, the docs are real, the commits are public, and the refund policy is unconditional. Let the product speak for itself.
Ship your next AI app in hours, not weeks.
Join the waitlist. Early birds get 50% off and first access to the alpha.
No spam. One email when the alpha ships.