# Agent Experience Score

> How well AI coding agents integrate Auth0 — measured across real-world tasks, not synthetic benchmarks.

**Average without Auth0 tools**: 71%
**Average with Auth0 tools**: 98%
**Models tested**: 8 (Claude Haiku 4.5, Claude Opus 4.8, Claude Sonnet 5, GPT-5.6 Luna, GPT-5.6 Sol, GPT-5.6 Terra, Gemini 3.1 Pro, Gemini 3.5 Flash)
**Frameworks tested**: 13 (Android, Angular, Express, Express API, FastAPI, Fastify API, Flask, Next.js, Nuxt, React, SPA JS, Swift, Vue)
**Total configurations**: 104

[Methodology](https://auth0.com/docs/get-started/auth0-agent-experience)

## Scores

| # | Model | Framework | Without Tools | With Auth0 Tools | Grade | Cost | Time |
|---|-------|-----------|---------------|------------------|-------|------|------|
| 1 | GPT-5.6 Luna | React | 100% | 100% | A | $0.17 | 35s |
| 2 | GPT-5.6 Terra | React | 100% | 100% | A | $0.33 | 32s |
| 3 | Claude Opus 4.8 | SPA JS | 100% | 100% | A | $0.46 | 52s |
| 4 | Claude Opus 4.8 | React | 100% | 100% | A | $0.62 | 1m 12s |
| 5 | Claude Opus 4.8 | Vue | 100% | 100% | A | $0.63 | 1m 28s |
| 6 | Claude Sonnet 5 | Vue | 100% | 100% | A | $0.67 | 2m 1s |
| 7 | Gemini 3.1 Pro | React | 100% | 100% | A | $0.72 | 59s |
| 8 | Claude Sonnet 5 | Angular | 100% | 100% | A | $0.74 | 2m 25s |
| 9 | Gemini 3.1 Pro | Angular | 100% | 100% | A | $0.74 | 1m 16s |
| 10 | Gemini 3.5 Flash | Vue | 100% | 100% | A | $0.76 | 1m 46s |
| 11 | Gemini 3.1 Pro | Vue | 100% | 100% | A | $0.92 | 1m 14s |
| 12 | Gemini 3.5 Flash | SPA JS | 93% | 100% (+7%) | A | $0.43 | 1m 4s |
| 13 | Gemini 3.1 Pro | SPA JS | 93% | 100% (+7%) | A | $0.50 | 51s |
| 14 | GPT-5.6 Terra | Vue | 92% | 100% (+8%) | A | $0.45 | 36s |
| 15 | Gemini 3.5 Flash | Express | 80% | 100% (+20%) | A | $0.39 | 59s |
| 16 | Gemini 3.1 Pro | Express API | 80% | 100% (+20%) | A | $0.44 | 55s |
| 17 | Claude Opus 4.8 | Next.js | 79% | 100% (+21%) | A | $0.62 | 1m 44s |
| 18 | Claude Sonnet 5 | Next.js | 79% | 100% (+21%) | A | $0.91 | 2m 46s |
| 19 | GPT-5.6 Terra | Nuxt | 65% | 100% (+35%) | A | $0.36 | 42s |
| 20 | Gemini 3.1 Pro | Next.js | 57% | 100% (+43%) | A | $0.95 | 1m 50s |
| 21 | Gemini 3.5 Flash | Next.js | 57% | 100% (+43%) | A | $1.39 | 2m 40s |
| 22 | Gemini 3.1 Pro | Android | 56% | 100% (+44%) | A | $2.36 | 2m 46s |
| 23 | Claude Sonnet 5 | Flask | 50% | 100% (+50%) | A | $0.60 | 1m 46s |
| 24 | Claude Sonnet 5 | FastAPI | 43% | 100% (+57%) | A | $0.30 | 47s |
| 25 | Gemini 3.1 Pro | FastAPI | 43% | 100% (+57%) | A | $0.39 | 40s |
| 26 | Claude Haiku 4.5 | Fastify API | 38% | 100% (+62%) | A | $0.08 | 37s |
| 27 | Gemini 3.5 Flash | FastAPI | 36% | 100% (+64%) | A | $0.47 | 1m 14s |
| 28 | GPT-5.6 Terra | Flask | 36% | 100% (+64%) | A | $0.57 | 42s |
| 29 | Gemini 3.5 Flash | Nuxt | 35% | 100% (+65%) | A | $0.91 | 1m 54s |
| 30 | Gemini 3.1 Pro | Fastify API | 31% | 100% (+69%) | A | $0.38 | 44s |
| 31 | Gemini 3.1 Pro | Nuxt | 30% | 100% (+70%) | A | $1.66 | 1m 55s |
| 32 | Gemini 3.5 Flash | Fastify API | 23% | 100% (+77%) | A | $0.67 | 1m 38s |
| 33 | GPT-5.6 Luna | Flask | 21% | 100% (+79%) | A | $0.23 | 58s |
| 34 | Gemini 3.1 Pro | Flask | 21% | 100% (+79%) | A | $0.79 | 1m 22s |
| 35 | GPT-5.6 Luna | Angular | 100% | 99% (-1%) | A | $0.19 | 45s |
| 36 | GPT-5.6 Terra | Angular | 100% | 99% (-1%) | A | $0.44 | 39s |
| 37 | Gemini 3.5 Flash | Angular | 100% | 99% (-1%) | A | $0.75 | 1m 39s |
| 38 | Gemini 3.5 Flash | React | 100% | 99% (-1%) | A | $0.93 | 1m 47s |
| 39 | GPT-5.6 Luna | Next.js | 86% | 99% (+13%) | A | $0.24 | 1m 4s |
| 40 | GPT-5.6 Terra | SPA JS | 86% | 99% (+13%) | A | $0.43 | 33s |
| 41 | GPT-5.6 Sol | Express API | 80% | 99% (+19%) | A | $1.61 | 1m 12s |
| 42 | GPT-5.6 Terra | Next.js | 79% | 99% (+20%) | A | $0.54 | 45s |
| 43 | GPT-5.6 Sol | FastAPI | 71% | 99% (+28%) | A | $0.79 | 52s |
| 44 | GPT-5.6 Luna | Swift | 70% | 99% (+29%) | A | $0.29 | 1m 2s |
| 45 | Claude Sonnet 5 | Android | 67% | 99% (+32%) | A | $0.67 | 2m 12s |
| 46 | Claude Haiku 4.5 | Next.js | 57% | 99% (+42%) | A | $0.20 | 1m 39s |
| 47 | Gemini 3.5 Flash | Android | 56% | 99% (+43%) | A | $0.64 | 1m 22s |
| 48 | Claude Opus 4.8 | Android | 56% | 99% (+43%) | A | $0.85 | 1m 50s |
| 49 | Gemini 3.1 Pro | Express | 47% | 99% (+52%) | A | $0.69 | 1m 10s |
| 50 | GPT-5.6 Terra | Fastify API | 46% | 99% (+53%) | A | $0.42 | 35s |
| 51 | Claude Haiku 4.5 | Android | 33% | 99% (+66%) | A | $0.27 | 1m 43s |
| 52 | Gemini 3.5 Flash | Flask | 29% | 99% (+70%) | A | $0.86 | 1m 47s |
| 53 | GPT-5.6 Luna | SPA JS | 100% | 98% (-2%) | A | $0.14 | 37s |
| 54 | Claude Sonnet 5 | React | 100% | 98% (-2%) | A | $0.47 | 2m 39s |
| 55 | GPT-5.6 Sol | Vue | 100% | 98% (-2%) | A | $0.93 | 59s |
| 56 | GPT-5.6 Sol | Angular | 100% | 98% (-2%) | A | $1.15 | 1m 20s |
| 57 | Claude Haiku 4.5 | Angular | 92% | 98% (+6%) | A | $0.58 | 5m 59s |
| 58 | GPT-5.6 Sol | SPA JS | 86% | 98% (+12%) | A | $0.99 | 1m 3s |
| 59 | GPT-5.6 Sol | Next.js | 86% | 98% (+12%) | A | $1.07 | 1m 16s |
| 60 | GPT-5.6 Terra | Express API | 80% | 98% (+18%) | A | $0.48 | 34s |
| 61 | GPT-5.6 Terra | Express | 80% | 98% (+18%) | A | $0.50 | 42s |
| 62 | GPT-5.6 Terra | FastAPI | 79% | 98% (+19%) | A | $0.33 | 30s |
| 63 | GPT-5.6 Luna | FastAPI | 71% | 98% (+27%) | A | $0.25 | 47s |
| 64 | GPT-5.6 Terra | Swift | 70% | 98% (+28%) | A | $0.63 | 49s |
| 65 | GPT-5.6 Terra | Android | 67% | 98% (+31%) | A | $1.34 | 1m 11s |
| 66 | GPT-5.6 Sol | Fastify API | 62% | 98% (+36%) | A | $1.57 | 58s |
| 67 | GPT-5.6 Luna | Android | 44% | 98% (+54%) | A | $0.25 | 50s |
| 68 | GPT-5.6 Luna | Fastify API | 31% | 98% (+67%) | A | $0.20 | 47s |
| 69 | GPT-5.6 Luna | Nuxt | 30% | 98% (+68%) | A | $0.38 | 1m 2s |
| 70 | Claude Haiku 4.5 | React | 100% | 97% (-3%) | A | $0.16 | 1m 2s |
| 71 | GPT-5.6 Luna | Vue | 100% | 97% (-3%) | A | $0.18 | 48s |
| 72 | GPT-5.6 Sol | React | 100% | 97% (-3%) | A | $1.58 | 1m 2s |
| 73 | Claude Haiku 4.5 | SPA JS | 93% | 97% (+4%) | A | $0.13 | 46s |
| 74 | Claude Sonnet 5 | SPA JS | 93% | 97% (+4%) | A | $0.44 | 1m 18s |
| 75 | GPT-5.6 Luna | Express | 87% | 97% (+10%) | A | $0.30 | 48s |
| 76 | GPT-5.6 Sol | Express | 87% | 97% (+10%) | A | $0.80 | 1m 1s |
| 77 | Claude Sonnet 5 | Express | 80% | 97% (+17%) | A | $0.47 | 1m 37s |
| 78 | Claude Sonnet 5 | Express API | 80% | 97% (+17%) | A | $0.52 | 1m 27s |
| 79 | Gemini 3.5 Flash | Express API | 80% | 97% (+17%) | A | $1.07 | 1m 59s |
| 80 | GPT-5.6 Sol | Android | 78% | 97% (+19%) | A | $2.41 | 1m 49s |
| 81 | GPT-5.6 Luna | Express API | 70% | 97% (+27%) | A | $0.30 | 58s |
| 82 | Gemini 3.5 Flash | Swift | 70% | 97% (+27%) | A | $0.62 | 1m 11s |
| 83 | Claude Opus 4.8 | Swift | 70% | 97% (+27%) | A | $0.73 | 1m 20s |
| 84 | Gemini 3.1 Pro | Swift | 70% | 97% (+27%) | A | $1.11 | 1m 22s |
| 85 | GPT-5.6 Sol | Nuxt | 70% | 97% (+27%) | A | $1.95 | 1m 45s |
| 86 | Claude Sonnet 5 | Swift | 60% | 97% (+37%) | A | $0.57 | 1m 34s |
| 87 | Claude Opus 4.8 | Angular | 100% | 96% (-4%) | A | $0.95 | 2m 4s |
| 88 | Claude Opus 4.8 | Express API | 80% | 96% (+16%) | A | $0.57 | 1m 12s |
| 89 | Claude Opus 4.8 | Fastify API | 77% | 96% (+19%) | A | $0.47 | 1m 1s |
| 90 | Claude Haiku 4.5 | Express | 73% | 96% (+23%) | A | $0.11 | 53s |
| 91 | Claude Opus 4.8 | FastAPI | 57% | 96% (+39%) | A | $0.58 | 1m 8s |
| 92 | Claude Opus 4.8 | Nuxt | 55% | 96% (+41%) | A | $0.94 | 1m 39s |
| 93 | Claude Sonnet 5 | Nuxt | 45% | 96% (+51%) | A | $1.05 | 2m 54s |
| 94 | Claude Haiku 4.5 | FastAPI | 43% | 96% (+53%) | A | $0.09 | 37s |
| 95 | GPT-5.6 Sol | Flask | 36% | 96% (+60%) | A | $2.11 | 1m 57s |
| 96 | Claude Opus 4.8 | Express | 80% | 95% (+15%) | A | $0.55 | 1m 18s |
| 97 | Claude Haiku 4.5 | Flask | 43% | 95% (+52%) | A | $0.12 | 56s |
| 98 | Claude Haiku 4.5 | Vue | 100% | 94% (-6%) | A | $0.24 | 1m 38s |
| 99 | Claude Haiku 4.5 | Swift | 70% | 94% (+24%) | A | $0.16 | 52s |
| 100 | GPT-5.6 Sol | Swift | 70% | 92% (+22%) | A | $4.21 | 2m 45s |
| 101 | Claude Sonnet 5 | Fastify API | 69% | 92% (+23%) | A | $0.34 | 1m 16s |
| 102 | Claude Opus 4.8 | Flask | 29% | 91% (+62%) | A | $0.66 | 1m 21s |
| 103 | Claude Haiku 4.5 | Nuxt | 40% | 89% (+49%) | B | $0.48 | 4m 27s |
| 104 | Claude Haiku 4.5 | Express API | 60% | 87% (+27%) | B | $0.42 | 4m 11s |

## Methodology

### 1. Real integration tasks
We test each model with realistic developer prompts — the kind you'd actually type when building an app. No optimized instructions, no guided walkthroughs.

### 2. Automated grading
Every response is scored across multiple dimensions: correct SDK usage, proper configuration, security best practices, and more. Grading is deterministic and reproducible.

### 3. Tool comparison
Each model runs twice: once with just its training knowledge (baseline), and once with Auth0's MCP Server and Agent Skills enabled. The delta measures the real-world impact of Auth0's developer platform.

## Developer Tools

### Auth0 MCP Server
Connect AI agents directly to Auth0 for real-time tenant management.

[Documentation](https://auth0.com/docs/get-started/build-with-ai-tools#auth0-docs-mcp-server)

### Auth0 Agent Skills
Pre-built implementation knowledge for AI coding agents.

[GitHub Repository](https://github.com/auth0/agent-skills)

## Grade Scale

- **A** (90-100%): Production-ready implementation
- **B** (75-89%): Good with minor issues
- **C** (60-74%): Functional but needs review
- **D** (<60%): Significant issues

*Last updated: July 2026*