Better Auth vs Clerk vs Auth.js in 2026: Which Auth for Your SaaS?

Auth.js joined Better Auth in Sep 2025; Clerk's free tier covers 50,000 retained users. We run Better Auth in 3 SaaS starters: cost, ownership, DX, lock-in.

SoarStarter Team

The auth landscape for TypeScript apps changed more in the last fourteen months than in the five years before. In September 2025, Auth.js (NextAuth) became part of Better Auth. The Auth.js team now recommends Better Auth for new projects, and Auth.js gets security and urgent fixes only. In July 2026, Better Auth joined Vercel. Meanwhile Clerk kept growing as the polished, hosted option. If your mental model is still "NextAuth or Clerk," it's out of date.

We run Better Auth in three full-featured SaaS starters: our Next.js, Nuxt and TanStack Start templates. We explained the original decision in Why We Chose Better Auth Over NextAuth and Clerk. This post is the 2026 head-to-head, with current versions, pricing and status.

Quick verdict

Better AuthClerkAuth.js (NextAuth)
ModelOpen-source library (MIT), runs in your appHosted service with SDKs and UI componentsOpen-source library
Status (Nov 2026)1.7.x, active; joined Vercel Jul 2026Commercial, activePart of Better Auth since Sep 2025; security and urgent fixes
Latest stable1.7 (Aug 2026)Hosted (SDKs versioned per framework)v4 (next-auth@4.24); v5 still in beta
Where users liveYour database (Drizzle, Prisma, Kysely, …)Clerk's infrastructureYour database, or stateless JWT
CostFreeFree to 50,000 monthly retained users per app, then Pro from $25/mo + $0.02 per retained userFree
Pre-built UINo (bring your own forms)Yes: sign-in, user profile, org switcherMinimal default pages
FrameworksFramework-agnostic: Next.js, Nuxt, TanStack Start, SvelteKit, Hono, …SDKs for Next.js, React, Nuxt, TanStack Start, Expo, …Next.js first; other adapters exist
Best fitTeams that want to own auth and extend it with pluginsTeams that want auth done in an afternoon and are fine renting itExisting apps already on it

Short answer: for a new self-hosted TypeScript SaaS in 2026, start with Better Auth. Choose Clerk when speed to a polished login, organizations UI and zero auth maintenance are worth a per-user bill and a hosted dependency. Choose Auth.js only if you already run it. Its own maintainers now point new projects to Better Auth.

What happened to Auth.js

Auth.js (the project behind NextAuth.js) was the default for Next.js apps for years. On September 22, 2025, Better Auth announced that Auth.js is "now part of Better Auth," and the Auth.js team posted the same news on their GitHub on September 26. Two points from those posts matter for your decision:

  • Existing apps are fine. The Auth.js team says it "continues to receive security and critical fixes." Better Auth still publishes Auth.js security updates (the most recent advisory we saw was July 2026).
  • New projects should start elsewhere. Both announcements recommend Better Auth for new projects. The one gap Better Auth calls out is stateless sessions without a database, which Auth.js does well.

There's also the version problem. next-auth@latest is still v4 (4.24.15, July 2026), while v5 has been in beta for years (5.0.0-beta.32). Starting a new app on a beta-forever major that is now in maintenance mode is hard to justify.

Better Auth: a library you own

Better Auth is a TypeScript auth framework that runs inside your app and stores everything in your database. It's MIT-licensed and free. Its July 2026 announcement about joining Vercel says Vercel shares its commitment to keeping auth "open source, framework and platform agnostic."

In our three web templates, the Better Auth setup is nearly identical across frameworks:

  • Email and password with required email verification, plus GitHub, Google and Sign in with Apple (Apple turns on when you set a bundle ID), with account linking.
  • Plugins: admin() for roles and the admin panel, emailOTP() for one-time codes (verification, passwordless sign-in, password reset), and bearer({ requireSignature: true }) so native apps can use the same backend with signed bearer tokens.
  • Rate limits on the sensitive endpoints (sign-in, sign-up, OTP and password reset).
  • Drizzle adapter: PostgreSQL on Next.js and Nuxt, Cloudflare D1 (SQLite) on TanStack Start.

The framework-specific part is tiny. TanStack Start adds one tanstackStartCookies() plugin, Nuxt mounts the handler as a Nitro route, and Next.js uses a catch-all route handler. That's the main reason we standardized on it: one auth model across three frameworks (Next.js auth docs, Nuxt, TanStack Start).

The cost of owning auth is that you build the screens and you're responsible for operating it: email delivery, secret rotation, database backups. Our templates ship the login, register, verification and reset pages, so that part is done if you start from one.

Clerk: auth as a service

Clerk is the fastest way to a beautiful, complete auth experience. Drop in <SignIn />, and you get social login, MFA, passkeys, user profile management, and B2B organizations with an org switcher (some of these on paid plans), all hosted, all maintained by someone else.

As of November 2026, Clerk's pricing bills on monthly retained users (MRUs), meaning users who come back at least a day after signing up, not on every sign-up. The Hobby plan is free up to 50,000 MRUs per app. Pro is $25/month ($20/month billed annually), includes 50,000 MRUs, and then charges $0.02 per MRU, falling in volume tiers. Business starts at $300/month. That's generous for a new product, but the bill grows with your success, and some features (like extra SSO connections) are priced separately.

The deeper trade-off is ownership. Your user records live on Clerk's infrastructure, and every sign-in depends on their uptime. Migrating off later means exporting users and rewriting every auth touchpoint. For many teams that's a fair price for never thinking about auth. For a template whose promise is "you own the code and deploy it anywhere," it isn't, which is why our starters don't use it.

Developer experience compared

  • Type safety: Better Auth infers the session type, including fields added by plugins, so session.user.role is typed after you add admin(). Clerk's SDKs are well typed too. Auth.js typing usually means augmenting module declarations by hand.
  • Extending: Better Auth's plugin system covers 2FA, passkeys, organizations, API keys and more, and plugins add their own tables and endpoints. Clerk extends through dashboard settings, metadata and webhooks. Auth.js extends through callbacks.
  • Multi-framework: Better Auth and Clerk both cover the major frameworks. Auth.js is still most at home in Next.js.

How to choose

  • Choose Better Auth if you want users in your own database, no per-user pricing, one auth model across frameworks, and room to grow through plugins. It's our default for every web template.
  • Choose Clerk if you need polished auth UI and B2B organizations this week, prefer paying to maintaining, and accept a hosted dependency for user identity.
  • Stay on Auth.js only if an existing app already runs it well, especially with stateless JWT sessions. Plan your migration to Better Auth on your own schedule.

Want to see Better Auth wired into a full SaaS (admin panel, Creem payments, email, i18n)? Read the authentication docs, try the live demo at next.soarstarter.com, or compare the Next.js, Nuxt and TanStack Start templates.