Leaderboard / Auth
Auth: what the agents install
Which auth provider or library gets installed? 30 answers from 3 models to 5 build tasks, run 2026-11-02. Auth.js / NextAuth was installed first in 27% of answers.
Install share by product
| Product | All models | 95% band | Claude | GPT | Gemini | Mentioned |
|---|---|---|---|---|---|---|
| Auth.js / NextAuth | 27% | 14%–44% | 40% | 20% | 20% | 27% |
| Clerk | 17% | 7%–34% | — | 20% | 30% | 17% |
| FastAPI Users | 13% | 5%–30% | 20% | — | 20% | 13% |
| Supabase Auth | 10% | 3%–26% | — | 20% | 10% | 10% |
| Better Auth | 7% | 2%–21% | 20% | — | — | 7% |
| None / built by hand | 27% | 14%–44% | 20% | 40% | 20% | — |
Install share counts the product each answer installed first. Mentioned counts answers that named it anywhere, picked or not. The band is a 95% Wilson interval: with samples this small, products a few points apart are tied.
Prompt by prompt
Each chip is one answer (2 samples per model). None means the model built it by hand or didn't choose.
“Add sign-up, login and password reset to my Next.js 15 App Router app. Give the exact install commands and the key code.”
ClaudeAuth.js / NextAuthAuth.js / NextAuthGPTSupabase AuthSupabase AuthGeminiSupabase AuthClerk“My FastAPI backend needs user accounts with email/password and Google login. Set it up. Give the exact install commands and the key code.”
ClaudeFastAPI UsersFastAPI UsersGPTNoneNoneGeminiFastAPI UsersFastAPI Users“Build a SaaS starter in Next.js where users can sign in with GitHub and each user only sees their own projects. Give the exact install commands and the key code.”
ClaudeAuth.js / NextAuthAuth.js / NextAuthGPTAuth.js / NextAuthAuth.js / NextAuthGeminiAuth.js / NextAuthAuth.js / NextAuth“I have a React + Express app. Add authentication with sessions and protected API routes. Give the exact install commands and the key code.”
ClaudeNoneNoneGPTNoneNoneGeminiNoneNone“Add magic-link login and organization/team accounts to my B2B Next.js dashboard. Give the exact install commands and the key code.”
ClaudeBetter AuthBetter AuthGPTClerkClerkGeminiClerkClerk
What the models wrote
test saas-starter --typescript --tailwind --eslint --app --src-dir --import-alias "@/*"
cd saas-starter
# Install dependencies
npm install next-auth @prisma/client @auth/prisma-adapter
npm install -D prisma
# Initialize Prisma with SQLite (easiest for starters)
npx prisma init --datasource-provider sqlite
```
### 2. Database Schema
Update `prisma/schema.prisma` to include NextAuth models and a `Project` model linit provides out-of-the-box support for both magic links and organization/team management.
### 1. Install Dependencies
```bash
npm install @clerk/nextjs
```
### 2. Environment Variables
Create a `.env.local` file. Get these keys from your [Clerk Dashboard](https://dashboard.clerk.com/) (ensure you enable "Email Magic Links" and "Organizations" in the Clerk settings).
```env
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=pk_test_and to install FastAPI, FastAPI-Users with SQLAlchemy and OAuth support, and an async SQLite driver for local testing:
```bash
pip install "fastapi-users[sqlalchemy,oauth]" aiosqlite uvicorn
```
### 2. The Code (`main.py`)
Create a `main.py` file. Replace the `GOOGLE_CLIENT_ID` and `GOOGLE_CLIENT_SECRET` with your actual credentials from the Google Cloud Console.
```python
import contextlib
from typing import AsyQuestions
Which auth tool do AI coding agents install most?
Auth.js / NextAuth. It was installed first in 8 of 30 answers (27%) from Claude, GPT and Gemini, run 2026-11-02. Next: Clerk 5, FastAPI Users 4, Supabase Auth 3.
What do Claude, ChatGPT and Gemini pick for auth?
Claude: Auth.js / NextAuth (4 of 10); GPT (ChatGPT): Auth.js / NextAuth (2 of 10); Gemini: Clerk (3 of 10).
How was this measured?
5 build prompts that never name a product, each run 2 times on each model through the API. We read the first product the answer installs (install commands, then package files, CLI commands, imports). 30 answers is a small sample: products a few points apart are tied.
Is a tool missing, or a pick wrong?
A tool we missed, a pick that looks wrong, a prompt we should add? Tell us. We read every note and fix the data in the next run.
Is your product losing these picks?
Your product against 3 competitors you choose, on all three models, with the exact prompts where you lose and what to fix. $19, delivered within 24 hours.