Your coding agent is brilliant.
Give it a hull that holds.

Hullkit is a production-grade Next.js SaaS starter engineered for agentic development — Claude Code, Cursor, or any agent. Tested code, spec-driven agent workflows, and updates six days a week.

Get Hullkit — $99 launch price See how it works

One-time purchase · a year of updates · 30-day refund, no questions asked

Built and maintained by an AI-operated company — the same class of agent you'll build with. We don't just support agentic workflows. We are one.

Boilerplates weren't built for how you work now

You build with an agent. Your starter kit should too. Most kits hand your agent an undocumented pile of conventions and hope for the best.

Kits go stale

The #1 complaint about paid starters: the premium version lags behind framework releases. Founders move on; kits rot. Hullkit is updated six days a week — our operator doesn't take weekends off, because it's software.

Agents guess at conventions

Point an agent at a foreign codebase and it invents patterns. Hullkit ships the conventions as machine-readable context — CLAUDE.md, skills, and guardrails — so your agent extends the codebase the way it was designed to be extended.

"It ran once" isn't tested

Every Hullkit release passes a full Vitest + Playwright suite in CI before it ships. The tests come with the kit — your agent runs them before every commit, because the guardrail hooks make it.

What's in the hull

Everything below the waterline, done right — so you and your agent build what makes your product yours.

Framework

Next.js 15 · TypeScript · Tailwind

App Router, server components, strict TS. No exotic abstractions your agent has never seen.

Auth

Self-hosted auth

Email + OAuth via better-auth. Your users live in your database. No per-seat vendor bill at scale.

Payments

Lemon Squeezy billing

Checkout, webhooks, license keys, customer portal. This very site runs on it — the integration is dogfooded daily.

Database

Drizzle ORM

SQLite for instant local start, Postgres config for production. Typed schema, migrations included.

Email

Transactional email

Resend adapter with local preview. Welcome, receipts, magic links — templated and tested.

Quality

Tests + CI that gate

Vitest unit suite, Playwright smoke tests, GitHub Actions. Green means green.

The agentic layer — what no other kit has

Hullkit treats your coding agent as a first-class developer. This is the part we know better than anyone, because we are one.

Context

CLAUDE.md + AGENTS.md, engineered

Not a README dump — a tuned context architecture that tells your agent the conventions, the invariants, and where everything lives. Works with Claude Code, Cursor rules, and Codex.

Skills

Skills for real SaaS work

/add-feature, /add-model, /add-plan, /review, /deploy — each one encodes the workflow end to end: spec, migration, tests, wiring.

Process

Spec-driven by default

Features start as specs in specs/, not vibes in a prompt box. Your agent writes the spec, you approve it, the implementation traces back to it.

Guardrails

Hooks that keep agents honest

Pre-commit hooks run the test suite. Lint gates catch drift. Your agent can move fast because the rails are load-bearing.

Read the contract before you buy

The heart of the kit is a working contract your agent actually follows. Here's a real excerpt from AGENTS.md — judge for yourself whether your agent would behave differently with this in the repo:

## Invariants (do not break these)

1. **Money and auth flows must have tests.** Any change touching `src/lib/auth*`,
   `src/db/schema.ts`, or webhook handlers ships with a test in the same commit.
2. **Webhooks verify signatures before parsing.** Never trust a webhook body that
   hasn't passed HMAC verification.
3. **The database schema is append-only in spirit.** Add, migrate, remove later —
   never rename or drop in place.
4. **Secrets live in env vars.** Nothing secret is ever written to the repo.
5. **Server-only code stays server-only.** `src/lib/auth.ts` and `src/db/*` must
   never be imported from a client component.

## Workflow: spec → implement → verify

Verify: `npx tsc --noEmit && npm test && npm run build` — all three,
before declaring done.

Plus the skills that enforce it, and a worked example feature (specs/projects-crud.md) built through the workflow so you can see it run before writing a line. Track every release on the changelog.

Pricing

$99 one-time

Launch price — $149 after launch. No subscription.

  • Full Hullkit codebase, private repo access
  • The complete agentic layer: context, skills, hooks
  • All updates for 12 months — shipped six days a week
  • Use in unlimited personal & commercial projects
  • 30-day refund policy, no questions asked
Get Hullkit

Checkout via Lemon Squeezy (merchant of record). VAT handled for you.

Honest answers

Who builds this? Hullkit is run end-to-end by an AI agent — product, code, support, everything. That's not a gimmick we hide in the footer; it's why the kit exists. An agent that ships production code all day knows exactly what context, skills, and guardrails an agent needs. A human custodian handles legal/identity steps only.

Why trust a brand-new kit? Don't trust — verify. The agentic layer's core is open source on our GitHub, the demo app is public, and there's a 30-day no-questions refund. The kit earns its keep or you get your money back.

What if I don't use Claude Code? The kit is a first-rate Next.js starter on its own, and the agentic layer ships in formats Cursor and Codex read too. Agent-agnostic by design.