Agent leaderboards / All sectors / Auth
Auth: which providers coding agents choose
Auth0 led auth with about 25% of 724 runs.
Read this leaderboard as textrankings, key learnings, method
Key learnings
We asked five coding agents to add auth to 17 codebases, 724 runs in all. We varied the wording each time and asked as four different people. No product came near half, and the agents wrote auth themselves in about 9% of runs.
Compliance runs went to a different provider
We framed 86 runs around procurement and compliance. Microsoft Entra ID won 44 of them, and Auth0 won six. In the plain ask, Auth0 took 174 of 638 runs.
Different wordings changed the pick
We put the same question to the same agent and codebase several times, in different words. In 63 of 85 of those groups, the runs did not all land on one product.
The agents did not agree on a leader
WorkOS AuthKit was the top pick for Claude Code, Cursor and Grok Build CLI. Auth0 led for Codex with 63 wins, and for Muse Code with 59.
Named often, chosen rarely or never
Auth.js came up in 163 runs and was never picked. Google Sign-In was named in 169 runs and won five. Supabase Auth was named in 168 and won once.
- The simulated user sent the agent back at least once in 120 runs.
- In 46 runs it refused to approve until the agent named a specific product.
- JWT came up in 409 runs and won none, but it is a token format these providers issue rather than a provider.
- Runs asked as a vibe coder topped out with Google Identity at 14 wins.
The ranking724 runs
| Product | Wins | Share | ||
|---|---|---|---|---|
| 1 | Auth0auth0.com | 180 | 25% | |
| 2 | WorkOS AuthKitworkos.com | 131 | 18% | |
| 3 | Clerkclerk.com | 86 | 12% | |
| 4 | Microsoft Entra IDentra.microsoft.com | 80 | 11% | |
| 5 | Built in-houseoutcome | 65 | 9% | |
| 6 | Better Authbetter-auth.com | 29 | 4% | |
| 7 | Amazon Cognitoaws.amazon.com | 23 | 3% | |
| 8 | Keycloakkeycloak.org | 22 | 3% | |
| 9 | Google Identitydevelopers.google.com | 14 | 2% | |
| 10 | Arcticarcticjs.dev | 11 | 2% | |
| 11 | Ory Kratosory.sh | 9 | 1% | |
| 12 | Laravel Fortify + Laravel Socialite | 8 | 1% | |
| 13 | Oktaokta.com | 8 | 1% | |
| 14 | Laravel Fortifylaravel.com | 7 | 1% | |
| 15 | django-allauthallauth.org | 6 | 1% | |
| 16 | ClassLink + Clever Instant Login | 5 | 1% | |
| 17 | Google Sign-Indevelopers.google.com | 5 | 1% | |
| 18 | Zitadelzitadel.com | 5 | 1% | |
| 19 | openid-clientgithub.com | 4 | 1% | |
| 20 | Remix Authgithub.com | 3 | 0% | |
| 21 | Microsoft Entra External IDentra.microsoft.com | 3 | 0% | |
| 22 | Google Cloud Identity Platformcloud.google.com | 3 | 0% | |
| 23 | Clever Instant Loginclever.com | 2 | 0% | |
| 24 | Azure AD B2Cazure.microsoft.com | 1 | 0% | |
| 25 | Laravel Jetstreamlaravel.com | 1 | 0% | |
| 26 | Authentikgoauthentik.io | 1 | 0% | |
| 27 | mozilla-django-oidcgithub.com | 1 | 0% | |
| 28 | Supabase Authsupabase.com | 1 | 0% | |
| 29 | Google Workspaceworkspace.google.com | 1 | 0% |
By agent, by persona, by wording
By agent
| Codex204 runs | Auth0 · 63then WorkOS AuthKit · 35 |
| Muse Code · Muse Spark 1.3204 runs | Auth0 · 59then Clerk · 26 |
| Cursor133 runs | WorkOS AuthKit · 27then Auth0 · 24 |
| Claude Code133 runs | WorkOS AuthKit · 40then Auth0 · 25 |
| Grok Build CLI · Grok 4.750 runs | WorkOS AuthKit · 16then Auth0 · 9 |
By persona
| Senior engineer243 runs | Auth0 · 90then Clerk · 41 |
| Junior developer223 runs | Auth0 · 80then WorkOS AuthKit · 61 |
| Enterprise team160 runs | Microsoft Entra ID · 76then WorkOS AuthKit · 32 |
| Vibe coder98 runs | Google Identity · 14then Arctic · 11 |
By what the ask stressed
| The plain ask638 runs | Auth0 · 174then WorkOS AuthKit · 119 |
| Procurement and compliance86 runs | Microsoft Entra ID · 44then WorkOS AuthKit · 12 |
A case is one codebase with one agent, asked several times in different words and as different people. 63 of 85 cases did not hold to a single provider.
How this was measured
Every number on this page comes from a controlled experiment. We took 17 small applications, asked 5 coding agents (Codex, Muse Code (Muse Spark 1.3), Cursor, Claude Code, Grok Build CLI (Grok 4.7)) to add auth to each of them, in several wordings and as a senior engineer and junior developer and enterprise team and vibe coder, and let the agent choose the product. Each run happened in a sandbox with the agent at a pinned version, and a judge read the session to record what was chosen. That is 724 runs. The interactive board shows every run with its session, its diff and the judge's verdict. A simulated user stood in for the owner of the codebase: it read the agent's plan and had to approve it before any code was written; it sent the agent back at least once in 120 runs. Read the methodology and the publications.
If you sell in this sector: what these numbers mean for a vendor.
Open the interactive boardThis page as MarkdownOther sectors
- Agent sandboxes
- Observability
- AI SRE
- Payments
- Deploy
- Email providers
- Product analytics
- Databases
- File storage
- LLM evals & observability
- Voice Agents
- Serverless functions
- Cloud
- AI gateway
- Bot protection
- Search
- Agent frameworks
- Performance in CI
- Document processing & OCR
- Usage-based billing
- Internationalization
- Security testing
- Maps
- Message queues
- AI search
- Code review
- E-signature
- Security platforms
- In-app chat & calls
- Vector search