Skip to the ranking
SaaS Starter Kits
10 kits assessed 13 weighted criteria 130 published scores

Best Next.js boilerplates, benchmarked for 2026

Ten starter kits taken apart and scored on the same thirteen criteria — B2B tenancy, B2C readiness, billing, LLM gateway, RAG, affiliates, self-hosting, tests and licence terms.

9.6/10

Boilerplate-Stack finishes first by 2.5 points — the widest margin this benchmark has recorded. It is the only kit that covers B2B and B2C, a multi-provider LLM gateway, a pgvector RAG knowledge base, a built-in affiliate programme and a Docker + Coolify self-hosting path in one codebase.

The short answers: best Next.js boilerplate by use case

If you already know what you are building, take the row that matches and skip the rest.

Best overall
Boilerplate-Stack Score 9.6
Wins or ties on ten of thirteen criteria. The only kit with a multi-provider LLM gateway that fails over, a pgvector RAG base, affiliates, an MRR/ARR console, a hardened supply chain and a release gate that tests RLS and all 43 admin pages — and the only one built to be vibe-coded.
Best for B2B teams
Boilerplate-Stack Score 9.6
Account-centric multi-tenancy with organisations, invitations and roles, then an admin console that reports MRR and ARR rather than just listing users — plus per-seat and credit billing and an affiliate programme for channel sales. Makerkit matches it on tenancy primitives, but ships nothing like the business console.
Best value
Boilerplate-Stack Score 9.6
$174 buys an LLM gateway, a RAG knowledge base, an affiliate system and a self-host path that would otherwise be months of work plus a stack of monthly subscriptions. Achromatic is the better buy strictly on licence terms — unlimited projects for $180.
Best for unlimited projects
Makerkit Score 7.1
The one axis the leader loses outright. Unlimited projects and lifetime updates, with the deepest tenancy primitives and the best auth score in the benchmark.
Best free option
MIT, actively maintained, with the strongest test tooling of any free kit here. You will write billing, tenancy and AI yourself.
Best for a weekend B2C launch
ShipFast Score 3.8
Scores low because the rubric weights what it deliberately omits. For one person shipping a single-tenant consumer app by Sunday, it is still the fastest route here.

The Next.js boilerplate ranking

Ordered by weighted score out of 10. Sort by any column — the scores do not change, only the order you read them in.

Next.js SaaS boilerplates ranked by weighted benchmark score. Prices are one-time unless stated; figures verified .
Licence
01
Boilerplate-Stack

B2B + B2C, multi-LLM gateway with failover, pgvector RAG, affiliates, credits, MRR/ARR console, EU AI Act, hardened supply chain, i18n, Coolify. Vibe-code ready.

9.6
$174from $290
1 project · 12 mo updates
02
Makerkit

Next.js 16 with Supabase, Drizzle or Prisma. Organisations, RBAC, super-admin, passkeys and MFA.

7.1
$349Teams $649
Unlimited · lifetime
03
supastarter

Next.js, Nuxt or TanStack Start on a Hono API. better-auth, Prisma or Drizzle, five payment providers.

7.0
€299Startup €799
Unlimited · lifetime
04
Achromatic

tRPC and Better Auth, both Prisma and Drizzle repos, Vercel AI SDK, per-seat and usage billing.

6.9
$180single tier
Unlimited · ongoing
05
PropelKit

Next.js 15 on Supabase with Inngest jobs, credits billing and 225 passing tests. Newest kit here.

6.5
$69from $219
Unlimited · 12 mo updates
06
Next.js BoilerplateFree

ixartz. Clerk, Drizzle, next-intl, and the strongest test tooling in the benchmark. 13.0k stars.

5.3
FreeMIT
Unlimited · open source
07
Bedrock

End-to-end types from Prisma through Pothos to typed React hooks. SSO, audit trails, SOC 2 logging.

5.1
$499enterprise
One-time purchase
08
Next.js SaaS StarterFree

Postgres, Drizzle, Stripe and a basic teams model. 16.0k stars, but last touched December 2025.

3.9
FreeMIT
Unlimited · open source
09
ShipFast

The category's best-known kit. Fast for single-tenant B2C; no multi-tenancy, RBAC, admin or tests.

3.8
$199from $299
Unlimited · lifetime
10
Shipixen

A codebase generator for marketing sites and MDX blogs. Scored against a rubric it was not built for.

3.4
$3791 yr $249
Unlimited builds · lifetime

Where the gap actually is: feature coverage compared

Most kits converge on the same five things: auth, Stripe, a dashboard, email and a marketing page. The 2.5-point gap at the top of this ranking comes from thirteen capabilities that a SaaS needs in year one and that almost nobody ships. Counts are out of the ten kits assessed.

A console for running the business

Not a user table — MRR and ARR on the overview, AI spend by tokens and cost, plus twelve working sections: subscriptions by plan, affiliate and referral reversal, licence revenue and expiry, job run history, error logs. Makerkit and Achromatic ship user-and-organisation admin panels; neither reports revenue.

1 of 10

Zero-retention AI routing with failover

Payloads sent with zero data retention, a circuit breaker in front, and automatic failover to another provider when one degrades. Every other kit that ships AI at all points at a single provider and hopes it stays up.

1 of 10

A hardened supply chain

pnpm only, a seven-day quarantine before new dependency releases are allowed in, lifecycle scripts off by default, signature verification at install, SHA-pinned CI and a zero-vulnerability gate that fails the build. Nobody else treats their dependency tree as an attack surface.

1 of 10

A release gate, not a test folder

A disposable Supabase stack seeded with four personas, pgTAP over schema, RLS and grants, Playwright across both dashboards and all 43 admin pages with real Stripe and SSE flows — then WCAG 2.2 in light and dark, three browsers, visual baselines and Lighthouse budgets. Everyone else ships a test folder.

1 of 10

LLM gateway across providers

OpenAI, Claude and Gemini behind one interface plus OpenRouter, with LangChain orchestration, SSE streaming and prompt caching. Achromatic and PropelKit ship a single-provider AI SDK; the rest ship nothing.

1 of 10

RAG knowledge base

Document ingestion for PDF, TXT and MD on pgvector with HNSW indexing, wired to the chat layer. Every other kit leaves you to choose a vector store and build the pipeline.

1 of 10

Affiliate & referral programme

Cash commissions with configurable tiers, recurring or one-time, and referral tracking in the same database as your billing. Everyone else expects Rewardful or Tolt — a monthly fee plus another webhook to maintain.

1 of 10

EU AI Act and GDPR, shipped not described

Article 50 implemented — AI-interaction notice, durable output marking, a public transparency page — plus the operator paperwork as templates: Article 30 records, an Article 35 DPIA scaffold, a breach runbook, a DPA register, an append-only consent trail and a 28-section Article 15 export. Retrofitting this into a live product is far worse than starting with it.

1 of 10

B2B and B2C in one codebase

Account-centric multi-tenancy with organisations, invitations and roles for business customers — and credit packs, single-user accounts and an installable PWA for consumers. Makerkit and Bedrock commit to B2B; ShipFast is B2C only.

1 of 10

Docker + Coolify self-hosting

A documented path onto your own VPS at a fixed monthly cost, instead of per-seat serverless pricing that scales with your success. Shipixen deploys anywhere but has no SaaS backend to host.

2 of 10

Cron jobs and background workers

Scheduled work as a first-class module — usage rollups, dunning, digest emails, index rebuilds. PropelKit ships Inngest and supastarter has a jobs layer; most kits expect you to wire a scheduler yourself on day three.

3 of 10

Credits, licences and multi-currency

Credit packs, licence keys, a Stripe customer portal, automated webhooks and multi-currency pricing. Achromatic matches most of this; PropelKit and supastarter cover part of it.

3 of 10

Vibe-code ready

A bundled agent workspace — rules, skills, instructions and subagents — so Claude Code or Cursor reads the architecture instead of guessing at it. Makerkit documents explicit agent support; most kits leave your agent to reverse-engineer their conventions from file names.

2 of 10

What you are actually buying: ten starter kits reviewed

The five kits worth a long look, and the two best-known ones that people buy for the wrong reasons.

9.6Rank 01

Every kit on this page ships auth and Stripe. This is the only one where the AI layer is load-bearing rather than a demo route: OpenAI, Claude and Gemini behind one interface — GPT-5.6, Claude Opus 5 and Gemini 3.6 — reachable through direct provider APIs or OpenRouter, behind a server-only boundary that sends zero-data-retention payloads and fails over to another provider when one degrades, with a circuit breaker in front. Add LangChain orchestration, SSE streaming, prompt caching, a multi-agent registry, token and cost tracking where one credit equals one token, and a knowledge base for PDF, TXT and MD documents on pgvector with HNSW indexing. Nothing else here comes close on that axis.

It also refuses the usual either-or. The account-centric multi-tenant model gives business customers organisations, invitations, roles and admin dashboards, while credit packs, licence keys, multi-currency pricing and an installable PWA cover consumer products from the same codebase. An affiliate programme with cash commissions and configurable tiers — recurring or one-time — is built in, where every competitor sends you to a paid third-party service.

Two more things the field mostly ignores. It ships a Docker and Coolify path, so you can run the whole stack on a VPS at fixed cost rather than watching a serverless bill scale with your success. And it implements EU AI Act Article 50 disclosure, which stops being optional the moment you sell an AI feature into Europe. Underneath: Next.js 16, React 19, strict TypeScript, Tailwind v4 and shadcn/ui.

The operational plumbing is there too, which is where most kits thin out. Scheduled cron jobs and background workers ship as first-class modules — the thing you need on day three for usage rollups, dunning emails, digest sends and index rebuilds, and the thing you would otherwise bolt on with a third-party scheduler. Add API keys, a notification centre, a media library and caching, and the day-two work is already done.

It is also the one kit here built to be vibe-coded. The repository ships an agent workspace — rules, skills, instructions and subagents — so Claude Code or Cursor reads the architecture instead of inferring it from file names. In practice that is the difference between an agent that extends your billing logic correctly and one that invents a second, parallel version of it three folders away.

Then there is the part most kits skip entirely: somewhere to actually run the business. The platform admin console opens on total users, organisations, active subscriptions and MRR and ARR, with AI spend broken out by requests, tokens, cost and top users across today, seven and thirty days. Behind it sit twelve working sections — users, organisations with member limits, subscriptions by plan, referrals and affiliates with manual reversal, licences with revenue and expiry, dynamic roles, CMS, background jobs with run history, error logs, platform settings and a multi-locale changelog.

That is the difference between an admin panel and an operations console. Most kits give you a table of users; this one lets you reverse a fraudulent affiliate conversion, extend a licence, watch a cron handler's run history and read the error that broke it, without opening a database client or shipping a deploy.

The security posture is the other outlier, and it runs past the usual OWASP checklist — CSP headers, route-scoped rate limiting, RLS policies, CSRF, input sanitisation and Turnstile — into the part nobody advertises: the supply chain. pnpm only, a seven-day quarantine before a new dependency release is allowed in, lifecycle scripts off by default, signature verification at install, SHA-pinned CI and a zero-vulnerability audit gate that fails the build. No other kit in this benchmark treats its own dependency tree as an attack surface.

Testing is where the gap becomes almost unfair. One command spins up a disposable Supabase stack, seeds member, admin, owner and platform-admin personas, proves the product end to end and destroys the database. pgTAP covers schema, RLS and grants — the layer most kits never test at all — while Playwright drives the public site, both dashboards and all 43 admin pages, plus real Stripe, chat/SSE and document flows. The same gate runs WCAG 2.2 A/AA in light and dark, overflow at 320/768/1280, a Chromium/Firefox/WebKit matrix, visual baselines and Lighthouse budgets, on top of 640 unit tests and 68 accessibility and responsive tests.

Compliance is shipped rather than described. Article 50 transparency is implemented — AI-interaction notice, durable output marking, a public transparency page — and the operator paperwork comes as templates: Article 30 records of processing, an Article 35 DPIA scaffold, a breach runbook, a subprocessor and DPA register, an append-only Article 7 consent trail and an Article 15 export running to 28 sections.

The catch is the licence. $174 buys one project and twelve months of updates. Makerkit, supastarter, Achromatic and ShipFast all sell unlimited projects with lifetime updates at a comparable price. That is the one axis where it finishes last of the paid kits, and it is the strongest argument against buying here if you plan to ship a portfolio of products from a single purchase.

Buy it if you are building a real SaaS — AI features, paying teams, an affiliate channel, EU customers — and you want those wired on day one. Look elsewhere if one licence has to cover several projects.

7.1Rank 02

If your customers are companies and you want the most-trodden path in the category, this is the safest purchase here. Multi-tenancy is the real thing — personal accounts and organisations, invitations, role-based permissions and a super-admin dashboard that works on arrival. Auth covers email, OAuth, passkeys and MFA, which is more than most kits attempt and the best score in this benchmark.

It is also the least likely to strand you. One licence gives three database paths — Supabase, Drizzle with Better Auth, or Prisma with Better Auth — and three frontends in Next.js 16, TanStack Start and React Router, on React 19 and Tailwind 4. Unlimited projects, lifetime updates.

What it does not do is AI, RAG or affiliates. There is no provider abstraction, no vector store, no commission system. If any of those are core to your product you are building them yourself on top.

Buy it if you are building B2B software with teams, roles and invitations, and you value lifetime updates across unlimited projects.

7.0Rank 03

The least opinionated kit that is still complete. supastarter separates its Hono API from the frontend, which is why it can ship Next.js, Nuxt and TanStack Start builds off one architecture. Almost every choice underneath is yours: Prisma or Drizzle, and five payment providers — Stripe, Lemon Squeezy, Polar, Creem and Dodo — rather than the usual Stripe-or-nothing.

Auth runs on better-auth with passkeys, 2FA, magic links and OAuth — the widest set on this page. Development is visibly active; the repository moved the day we checked it.

The trade is coherence, and the same AI gap as Makerkit. Optionality means more decisions before you write a feature, and the Hono split adds a layer a solo founder shipping something simple does not need.

Buy it if you want to keep database, payment and framework options open, or you sell outside Stripe's supported countries.

6.9Rank 04

The best price-to-substance ratio among the mid-table kits. $180 buys unlimited projects for one team and both the Prisma and Drizzle repositories — you pick where to start, not what to buy.

Its billing is the deepest outside the leader: Stripe subscriptions and one-time payments, a self-service portal, per-seat pricing, usage-based pricing and a prepaid credit system for AI features. Add tRPC, TanStack Query, Better Auth with TOTP 2FA, Sentry, Pino, S3 presigned uploads and Fumadocs, and very little is missing.

What is missing is internationalisation, any compliance posture beyond boilerplate legal pages, and a RAG layer — its AI is the Vercel AI SDK with a streaming chat UI rather than a gateway. It is also a smaller project, so the community you fall back on is correspondingly smaller.

Buy it if your budget is tight and your billing is complicated — especially per-seat or usage-based pricing.

6.5Rank 05

The surprise of this round. At $69 on launch pricing it lands within half a point of kits selling for five times as much: multi-tenancy with RBAC, credits and usage billing, Stripe invoicing, Inngest background jobs, phone OTP alongside the usual auth methods, and 225 passing tests — a number most paid kits cannot quote because they have none.

Treat the score with more caution than the others. This is the youngest project in the benchmark, and a boilerplate's real value shows up in its second year, when Next.js ships a breaking change and you find out whether anyone is still maintaining it. Twelve months of updates, not lifetime.

Buy it if the launch price is live and you are comfortable betting on a young project. Wait if you need a maintenance track record.

3.8Rank 09

The most recognised name in the category finishes ninth, and the reason is worth stating plainly rather than hiding behind a number. ShipFast has no multi-tenancy, no role-based permissions, no admin panel, no test suite, no AI layer and no affiliate system. The rubric assigns most of its weight to exactly those things.

That is a statement about the rubric as much as the product. ShipFast was built so one person can put a single-tenant consumer app in front of users in a weekend, and at that job it is excellent — it earns the highest B2C score here. Stripe or Lemon Squeezy, Google OAuth and magic links, MongoDB or Supabase, a large community, unlimited projects and lifetime updates for $199.

Two cautions independent of the score. The last update we could verify was around six months ago, slow for a category where the framework moves every quarter. And there is no upgrade path: when your consumer app acquires its first business customer who wants to invite a colleague, you are rewriting the data model, not extending it.

Buy it if the product is single-tenant, consumer-facing and you want it live this week. Do not buy it if "teams" appears anywhere on your roadmap.

3.4Rank 10

Last place, and the ranking is misleading — so read the sentence rather than the number. Shipixen is not a SaaS backend. It is a generator that produces a Next.js 15 marketing site and MDX blog from 63 themes and 37 landing-page components, with SEO metadata, sitemaps and one-click deployment handled for you.

Judged on auth, billing and tenancy it scores near zero, because it does not claim to do any of them. Judged on producing a fast, good-looking, well-optimised marketing site without writing CSS, it has no real competitor in this list.

Buy it if you need the marketing site around a product you are building elsewhere. It pairs with any kit above rather than replacing one.

How the scores are made

Thirteen criteria, each scored 0–10 from published documentation, public repositories and pricing pages, then weighted and summed to a mark out of 10. The weights encode one assumption: that you are building a commercial, AI-era SaaS in 2026 that intends to charge both businesses and individuals. If that is not your product, re-weight them — every sub-score is published below, so you can.

Billing & monetisation

Subscriptions, one-time, per-seat, usage metering, credits, licences, portal, multi-currency.

11%
AI & LLM gateway

Multi-provider abstraction, OpenRouter, orchestration, streaming, prompt caching, agents.

11%
Auth & security

Methods offered, passkeys and MFA, session handling, row-level security, OWASP posture.

9%
B2B multi-tenancy

Organisations, invitations, RBAC, ownership transfer, admin dashboards.

9%
Stack freshness

Next.js and React versions, Tailwind 4, ORM currency, and how recently the repo moved.

9%
Testing & QA

Unit and E2E coverage shipped, CI configuration, component tooling.

8%
RAG & knowledge base

Document ingestion, vector store, indexing strategy, retrieval wired to the chat layer.

7%
Affiliates & growth

Referral tracking, commission tiers, payouts — built in rather than bolted on.

7%
Agent readiness

Bundled rules, skills or docs that let Claude Code and Cursor infer the architecture.

7%
Deploy & self-host

Docker, Coolify or VPS paths as well as Vercel; how much hosting lock-in you inherit.

6%
i18n & compliance

Translation infrastructure, multi-currency, GDPR and EU AI Act provisions, audit logging.

6%
B2C readiness

Single-user accounts, consumer checkout, credit packs, PWA, time-to-first-launch.

5%
Licence value

Price against scope: projects permitted, seats, update window, redistribution terms.

5%

What the method cannot see

Scores are built from what each vendor publishes and what is visible in public repositories, not from running every kit in production for a quarter. Code quality inside paid private repositories is assessed indirectly, and support responsiveness is not scored at all. The leader is beaten or matched on seven of the thirteen criteria — auth, billing, B2B tenancy, B2C, stack, tests and licence value — and finishes last of the paid kits on licence terms; the margin comes from the six axes where the rest of the field ships nothing. Prices and terms were verified on and change often; check before you buy.

Every score, published

All 130 sub-scores, including the seven criteria where the leader is beaten or matched. Weighted total in the final column.

Scroll the table sideways to see every criterion.

Sub-scores out of 10 per criterion. Weighted total out of 10. Verified .
Starter kit Bill
11%
AI
11%
Auth
9%
B2B
9%
Stack
9%
Test
8%
RAG
7%
Affil
7%
Agent
7%
Host
6%
i18n
6%
B2C
5%
Lic
5%
Total
Boilerplate-Stack 101091010101010101010959.6
Makerkit 84101010922866797.1
supastarter 9410910722678797.0
Achromatic 1079996315647106.9
PropelKit 96888821763796.5
Next.js Boilerplate 5076910003787105.3
Bedrock 71987600368565.1
Next.js SaaS Starter 605772002616103.9
ShipFast 72615200262993.8
Shipixen 34317300383473.4

Which Next.js boilerplate you should buy

The ranking answers "which is strongest overall". This answers the question you actually have.

You are building a real SaaS to sell

Paying customers, an AI feature, teams, an affiliate channel, EU users — the full shape, not a prototype.

The only kit that covers all of it in one codebase: LLM gateway with OpenRouter, pgvector RAG, B2B tenancy and B2C credits, affiliates with commission tiers, Article 50 disclosure, and a Coolify self-host path.

Your product's core feature is AI

Chat, retrieval over customer documents, generation metered by usage — the AI is the thing people pay for.

Four providers behind one interface with LangChain orchestration and prompt caching, a pgvector knowledge base, and credit packs that meter real token spend. Achromatic at $180 is the fallback if a single-provider AI SDK is genuinely enough.

You want lifetime updates on unlimited projects

You ship several products a year and one licence has to cover them all.

Makerkit7.1 · $349

This is the one axis where the leader finishes last of the paid kits. Makerkit gives unlimited projects and lifetime updates with the deepest B2B tenancy and the best auth score in the benchmark.

You want to keep your options open

Undecided on database, payment provider or even framework — or selling where Stripe does not reach.

supastarter7.0 · €299

Prisma or Drizzle, five payment providers, and Next.js, Nuxt or TanStack Start from one architecture. Nothing else here gives you this many exits.

You are learning, or the budget is zero

A side project, a first SaaS, or you want to read every line before you trust it.

MIT, actively maintained, with Vitest, Playwright and Storybook already wired — you will learn more from its test setup than from most paid kits. Budget weeks for billing and tenancy.

You sell into regulated industries

SSO, audit trails and SOC 2 evidence are on the buyer's checklist before the first demo.

Bedrock5.1 · $499

It scores mid-table because it ignores AI, affiliates and RAG entirely, but it is the only kit here shipping SSO, audit trails and SOC 2 logging as first-class concerns.

Questions people ask before buying a starter kit

What is a Next.js boilerplate?

A pre-built codebase that ships the parts every SaaS needs before it can charge anyone: authentication, subscription billing, a database schema, transactional email and an application shell. You clone it, rebrand it, and start writing the feature that is actually your product. The work it replaces is roughly four to ten weeks of engineering that no customer will ever pay you for.

Is a Next.js boilerplate the same thing as a starter kit?

In this market, yes — the two words are used interchangeably and the label predicts nothing about what you actually get. Every kit in this benchmark calls itself one or the other, and the noun on the marketing page has no bearing on whether it ships multi-tenancy, billing or tests.

If there is a shade of difference, it is that "boilerplate" usually means a repository you clone and edit in place, while "starter kit" sometimes implies a generator that scaffolds a project for you — Shipixen is the one kit here that works that way. Judge them on the thirteen criteria above, not on which word the vendor picked.

Which Next.js boilerplate is the best in 2026?

Boilerplate-Stack, at 9.6 out of 10. It is the only kit in this benchmark covering B2B multi-tenancy and B2C single-user mode, a multi-provider LLM gateway with OpenRouter, a pgvector RAG knowledge base, a built-in affiliate programme, cron jobs and a Docker plus Coolify self-hosting path, and the only one that ships a vibe-code-ready agent workspace.

Makerkit follows at 7.1 and is the stronger choice if you want lifetime updates across unlimited projects, which is the one axis where the leader finishes last of the paid kits.

Which Next.js boilerplate is best for vibe coding?

Boilerplate-Stack. It is the only kit in this benchmark that ships an agent workspace in the repository — rules, skills, instructions and subagents — so Claude Code or Cursor reads the architecture instead of inferring it from file names.

That is the difference between an agent that extends your existing billing logic and one that invents a second, parallel version of it three folders away. Makerkit documents explicit agent support and is the next best option; most kits ship nothing.

Are paid Next.js boilerplates worth it?

If you are shipping a commercial product, almost always. The paid kits here cost between $69 and $499, against four to ten weeks of undifferentiated engineering.

The honest exception is learning. If you want to understand how Stripe webhooks or row-level security actually work, a free MIT-licensed starter you read line by line is worth more than a paid one you never open.

What is the best free Next.js SaaS starter?

Next.js Boilerplate by ixartz, at 5.3 — MIT-licensed, actively maintained, and its testing setup of Vitest, Playwright and Storybook is the strongest of any free kit here. It ships Clerk for auth and next-intl for translations.

What it does not ship is billing, multi-tenancy or any AI layer, so budget several weeks for those.

Makerkit vs Boilerplate-Stack — which should I buy?

This is the closest contest in the benchmark, and the only pairing where both kits genuinely cover B2B. Makerkit ships organisations, invitations, role-based permissions and a super-admin on Next.js 16 with Supabase and a choice of Drizzle or Prisma, and it is the stronger buy on two axes. Authentication is one: passkeys and MFA earn it a 10 against Boilerplate-Stack’s 9. Licensing is the other, and it is the bigger one — $349 buys unlimited projects with lifetime updates, against $174 for a single project and twelve months. If you build client work, or expect to ship several products from one purchase, that gap compounds.

The 2.5 points between them are almost entirely what sits above the tenancy layer. Boilerplate-Stack ships a multi-provider LLM gateway with failover, a pgvector RAG knowledge base, an affiliate programme with cash commissions, credit billing, pg_cron scheduled jobs with a webhook handler dashboard, and an admin console that reports MRR and ARR rather than listing users — twelve sections across 43 pages, each of them under test. Makerkit’s super-admin is real but narrower, and it expects you to bolt on a third-party service such as Rewardful or Tolt for affiliates.

Both are ahead of the field on vibe coding, though not equally: Makerkit documents explicit agent support, while Boilerplate-Stack ships the agent workspace inside the repository — rules, skills, instructions and subagents — so Claude Code or Cursor reads the architecture instead of inferring it from file names.

Buy Makerkit if unlimited projects, passkeys, or a non-Next.js build decide it for you — it also ships TanStack Start and React Router. Buy Boilerplate-Stack if you are building one commercial, AI-era product and want the billing, scheduling, admin and AI layers already wired rather than on your backlog.

Which boilerplate supports both B2B and B2C?

Boilerplate-Stack is the only kit here scoring 9 or above on both. Its account-centric multi-tenant model covers organisations, invitations and role-based permissions for business customers, while credit packs, licence keys, multi-currency pricing and an installable PWA cover consumers.

Most kits commit to one shape: Makerkit and Bedrock lean B2B, ShipFast is single-tenant B2C only.

Do any Next.js boilerplates include an affiliate programme?

Only one in this benchmark ships a usable one. Boilerplate-Stack includes an affiliate and referral system with cash commissions and configurable tiers, recurring or one-time, in the same database as your billing.

Every other kit assessed expects you to bolt on a third-party service such as Rewardful or Tolt — a monthly cost and another webhook integration to maintain.

Can I self-host a Next.js SaaS boilerplate instead of using Vercel?

Yes, and it is worth checking before you buy, because most kits assume Vercel. Boilerplate-Stack documents a Docker and Coolify path, which lets you run the whole stack on your own VPS for a fixed monthly cost rather than per-seat serverless pricing that scales with your success.

Shipixen deploys cleanly to any Next.js host. The rest will run in Docker with work, but the configuration is yours to write.

What if I am not using Next.js?

supastarter also ships Nuxt and TanStack Start builds from the same architecture, and Makerkit ships TanStack Start and React Router alongside Next.js — both are worth a look if you are still framework-shopping.

Outside the React ecosystem, Open SaaS is a strong MIT-licensed option, but it is built on Wasp rather than Next.js, so it sits outside this benchmark.

How often is this ranking updated?

Prices, licence terms and stack versions are re-checked quarterly, and any kit shipping a major release is re-scored within the same quarter. The verification date is printed on the ranking table so you can judge how fresh the figures are before trusting them.