# Resources

Guides d’examen officiels, inscription, documentation, cours gratuits et lectures de fond pour les quatre certifications.

## Official programme

| Ressource | URL |
| --- | --- |
| Anthropic Partner Academy (inscription, guides d’examen, cours de préparation, FAQ) | `https://anthropic-partners.skilljar.com/` |
| Page de certification CCAO-F | `https://anthropic-partners.skilljar.com/claude-certified-associate-foundations-certification` |
| Page de certification CCDV-F | `https://anthropic-partners.skilljar.com/claude-certified-developer-foundations-certification` |
| Page de certification CCAR-F | `https://anthropic-partners.skilljar.com/claude-certified-architect-foundations-certification` |
| Page de certification CCAR-P | `https://anthropic-partners.skilljar.com/claude-certified-architect-professional-certification` |
| Pearson VUE – programme Anthropic (planification, OnVUE, politique de reprise) | `https://www.pearsonvue.com/us/en/anthropic.html` |
| Claude Partner Network (adhésion gratuite) | `https://claude.com/partners` |
| Support certification (corrections de nom) | `certifications-support@anthropic.com` |

Téléchargez le **PDF de l’Exam Guide** courant pour votre examen depuis sa page Partner Academy avant de réserver ; les blueprints et les poids peuvent changer d’une version à l’autre (ce site reflète la v1.0, juillet 2026).

## Free training (Anthropic Academy on Skilljar)

`https://anthropic.skilljar.com/`

| Cours | Le plus pertinent pour |
| --- | --- |
| Claude 101 | CCAO-F |
| AI Capabilities and Limitations | CCAO-F |
| AI Fluency (role tracks) | CCAO-F |
| Building with the Claude API | CCDV-F, CCAR-F, CCAR-P |
| Introduction to Model Context Protocol | CCDV-F, CCAR-F |
| MCP: Advanced Topics | CCAR-F, CCAR-P |
| Claude Code in Action / Claude Code 101 | CCDV-F, CCAR-F, CCAR-P |
| Introduction to Agent Skills | CCDV-F, CCAR-F |
| Introduction to Subagents | CCDV-F, CCAR-F |
| Claude with Amazon Bedrock / Google Vertex AI | CCAR-P |

## Documentation

### Claude API and platform – `platform.claude.com/docs/en/`

- Models overview · Pricing · Model deprecations · Fable 5.1 migration guide
- Messages API · Streaming · Handling stop reasons · Refusals and fallback
- Tool use overview · Tool search tool · Programmatic tool calling · Memory tool · MCP connector
- Thinking · Preserved thinking · Effort · Task budgets · Mid-conversation system messages
- Prompt caching · Cache diagnostics · Context editing · Compaction
- Structured outputs · Citations · Vision · PDF support · Files API
- Message Batches · Rate limits · Errors
- Managed Agents overview · `ant` CLI · Admin API
- Prompt engineering guide

### Claude Code – `code.claude.com/docs/en/`

- Overview · CLI reference · Settings · Memory and CLAUDE.md
- Subagents · Skills · Commands · Hooks · Plugins · MCP
- Agent teams · Dynamic workflows · Routines
- Claude Agent SDK overview (`/agent-sdk/`)
- Changelog: `https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md`

### Model Context Protocol – `modelcontextprotocol.io`

- Introduction · Specification · Transports · Authorization
- Python SDK (`mcp`, FastMCP) · TypeScript SDK (`@modelcontextprotocol/sdk`)

## Background reading (Anthropic engineering and research)

| Titre | Pourquoi c’est important |
| --- | --- |
| Building Effective Agents | Le cadrage workflow-vs-agent et les six patterns que testent les examens Architect |
| Effective Context Engineering for AI Agents | Économie du contexte, compaction, isolation |
| Advanced Tool Use | Tool search, programmatic tool calling, qualité des descriptions |
| Anthropic Usage Policy | Usage approprié vs inapproprié (domaines de gouvernance Associate et Professional) |

## Community study material (unofficial)

| Ressource | Notes |
| --- | --- |
| `github.com/dnacenta/claude-certified-architect` | Deep dive CCAR-F activement maintenu avec des aperçus des examens frères ; mise à jour septembre 2026 |
| `github.com/SGridworks/claude-certified-architect-training` | Programme de formation de 12 semaines |

Le matériel communautaire est utile mais non officiel. Quand il diverge de la documentation en direct ou de l’Exam Guide courant, les sources officielles priment.

## How to verify facts live

```python
from anthropic import Anthropic
client = Anthropic()

for m in client.models.list():
    print(m.id, m.display_name)

info = client.models.retrieve("claude-opus-5")
print(info.max_input_tokens, info.max_tokens, info.capabilities)
```

Utilisez-le avant un examen pour confirmer la gamme, et en production plutôt que de coder les limites en dur.

## Annotated reading list

Lisez dans cet ordre ; chaque élément s’appuie sur le précédent. Les annotations disent *pourquoi* un élément mérite votre temps et *quel objectif d’examen* il sert.

| # | Lecture | Pourquoi elle mérite votre temps | Sert |
| --- | --- | --- | --- |
| 1 | **Building Effective Agents** (Anthropic) | Le cadrage canonique workflow-vs-agent et les six patterns composables. Presque chaque item Architect « quel pattern » remonte ici. Notez le conseil répété de préférer la conception la *plus simple*. | F, P (D1) |
| 2 | **Effective Context Engineering for AI Agents** | Le contexte comme ressource budgétée : compaction vs editing, isolation via subagents, pourquoi le contexte long n’est pas gratuit. Sous-tend les items de fiabilité. | F, P (D5) |
| 3 | **Advanced Tool Use** (Anthropic engineering) | Qualité des descriptions d’outils, tool search, `defer_loading`, programmatic tool calling, appels parallèles. Les réponses au « pourquoi n’appelle-t-il pas mon outil » vivent ici. | D, F (D4/D5) |
| 4 | Docs **Messages API + Handling stop reasons** | Le seul champ sur lequel chaque boucle branche ; l’ensemble complet des `stop_reason` et leur gestion correcte. | D, F (D1/D3) |
| 5 | Docs **Prompt caching + Context editing** | Le calcul du seuil de rentabilité, la disposition des breakpoints, l’élagage côté serveur pour les modèles append-only. | D, F, P |
| 6 | **Fable 5.1 migration guide** | Le cluster de breaking changes (outils forcés, liaison du thinking, append-only, rétention) qui alimente de nombreux distracteurs. | D, F, P (D2) |
| 7 | **MCP Specification + Authorization** | Primitives, transports, négociation de capacités, OAuth 2.1/PKCE, propagation d’identité. | D, F (D4) |
| 8 | Docs **Claude Code: Memory, Hooks, Subagents, Skills** | La hiérarchie de configuration et le principe d’application-par-hooks ; la décision Skill-vs-CLAUDE.md-vs-command-vs-subagent-vs-MCP. | F, P (D2) |
| 9 | **Anthropic Usage Policy** | Usage approprié vs inapproprié, divulgation, domaines à haut risque. | A, P (gouvernance) |
| 10 | **RAG / retrieval docs + reranking guidance** | Chunking, récupération hybride, reranking, séparation des evals récupération-vs-génération. | P (D3) |
| 11 | **Anthropic evals guidance** | Golden sets, rubric grading, LLM-as-judge en session séparée, A/B par paires, reporting par segment. | D, P (D4/D8) |

:::tip[Discipline de lecture]
Lisez pour la *règle de décision*, pas la prose. Après chaque élément, écrivez la « signal → choix » en une phrase que vous en avez extraite et ajoutez-la à votre revue du Lexique. Si vous ne pouvez énoncer la règle, relisez cette section.
:::

## Per-domain documentation deep links

Sautez directement à la doc qui régit chaque domaine du blueprint. Les chemins sont sous `platform.claude.com/docs/en/` (API) et `code.claude.com/docs/en/` (Claude Code) sauf indication contraire.

### CCAO-F (Associate – Foundations)

| Domaine | Docs / cours principaux |
| --- | --- |
| D1 Prompting & Task Execution | Prompt engineering guide ; Academy *Claude 101* |
| D2 Output Evaluation & Validation | *AI Capabilities and Limitations* ; guidance sur l’hallucination et la provenance |
| D3 Product & Model Selection | Models overview ; Pricing ; claude.ai vs Desktop vs Code |
| D4 Workflow Integration & Solution Design | *AI Fluency* ; Projects and Connectors overview |
| D5 Configuration & Knowledge Management | Projects, custom instructions, knowledge files ; Memory (product) |
| D6 Governance, Risk & Responsible Use | Usage Policy ; data handling ; disclosure guidance |
| D7 Troubleshooting & Optimization | Prompt engineering : iterating ; quand escalader vers un développeur |

### CCDV-F (Developer – Foundations)

| Domaine | Docs principaux |
| --- | --- |
| D1 Applications & Integration | Messages API ; Streaming ; Vision ; PDF ; Files API ; Errors ; Rate limits |
| D2 Model Selection & Optimization | Models ; Pricing ; Prompt caching ; Message Batches ; Effort ; Thinking |
| D3 Agents & Workflows | Tool use ; Building Effective Agents ; Subagents |
| D4 Prompt & Context Engineering | Prompt engineering guide ; Context editing ; Compaction |
| D5 Tools & MCPs | Tool use overview ; Tool search ; MCP intro ; MCP connector |
| D6 Security & Safety | Prompt injection ; Usage Policy ; permissions/hooks |
| D7 Claude Code | Claude Code overview ; CLI reference ; headless mode |
| D8 Eval, Testing & Debugging | Evals guidance ; structured outputs ; traces |

### CCAR-F (Architect – Foundations)

| Domaine | Docs principaux |
| --- | --- |
| D1 Agentic Architecture & Orchestration | Building Effective Agents ; Subagents ; Dynamic workflows ; Agent teams |
| D2 Claude Code Configuration & Workflows | Memory/CLAUDE.md ; Settings ; Hooks ; Skills ; Commands ; Plan mode ; Plugins |
| D3 Prompt Engineering & Structured Output | Prompt engineering ; Structured outputs ; `strict` ; validation-retry |
| D4 Tool Design & MCP Integration | Advanced Tool Use ; MCP Specification ; Authorization ; MCP inspector |
| D5 Context Management & Reliability | Context engineering ; Context editing ; Compaction ; stop reasons ; retries |

### CCAR-P (Architect – Professional)

| Domaine | Docs principaux |
| --- | --- |
| D1 Solution Design & Architecture | Building Effective Agents ; Managed Agents vs Agent SDK ; ADR practice |
| D2 Claude Models, Prompting & Context | Models ; Fable 5.1 migration ; Prompt caching ; Context editing |
| D3 Integration (RAG) | RAG/retrieval docs ; hybrid search ; reranking ; citations |
| D4 Evaluation, Testing & Optimization | Evals guidance ; A/B and significance ; CI regression ; observability |
| D5 Governance, Safety & Risk | Usage Policy ; GDPR/HIPAA/FedRAMP ; ZDR ; DPIA ; guardrail layering |
| D6 Stakeholder Communication & Lifecycle | Solution lifecycle ; decision matrices ; SLO/SLA ; runbooks |
| D7 Developer Productivity & Operational Enablement | Claude Code settings ; managed policy ; team onboarding ; plugins |

:::note[Vérifier en direct]
Les URL de documentation et les noms de cours changent. Si un lien renvoie une 404, cherchez dans la Partner Academy et `platform.claude.com/docs` plutôt que de vous fier à un chemin enregistré — et confirmez toujours les poids de l’Exam Guide courant avant de réserver.
:::
