Claude Architect Pro
Claude Certified Architect: Professional
An independent study guide for the CCAR-P exam blueprint: solution design, model and prompt engineering, integration, evaluation, governance, stakeholder work and developer enablement. Pitched at architects who already ship systems, and written from scratch. Not affiliated with, endorsed by or sponsored by Anthropic, and it reproduces no exam questions.
8 modules · 11 hr 13 min of reading
3 of 40 lessons are open to read without an account, and a free account opens the other 37. Creating one costs nothing: there is no paid plan, no tier and nothing to buy.
- lessons
- 40
- flashcards
- 200
- quiz questions
- 160
- practice questions
- 506
Practice exam
A sitting is 63 questions in 2 hr, reported on the 100 to 1000 scale the programme publishes, with 720 to pass.
- Drawn fresh from a pool of 506, allocated across the domains by the weights below
- Worked through unseen questions first, so a second sitting is a different paper
- Feedback held back until you submit, then every option explained and the score split by domain
- Solution Design and Architecture
- 17 per cent of the paper
- Claude Models, Prompting and Context Engineering
- 13 per cent of the paper
- Integration
- 19 per cent of the paper
- Evaluation, Testing and Optimisation
- 16 per cent of the paper
- Governance, Safety and Risk Management
- 14 per cent of the paper
- Stakeholder Communication and Lifecycle Management
- 14 per cent of the paper
- Developer Productivity and Operational Enablement
- 7 per cent of the paper
The weights are the published blueprint's. The questions are not: all 506 were written here from the published objectives, and none of them is a real exam question.
Course outline
Module 1
Open to readOrientation: what CCAR-P measures
What the Professional exam is, how it differs from Foundations, the seven domains and their weights, and how to prepare. Read this first: it also sets out what this guide is and is not.
3 lessons · 43 min
Before you start
- What the CCAR-P exam measures
- Professional versus Foundations: the change in level
- How to prepare: build one system, then defend it
Module 2
Free accountDomain 1: Solution design and architecture
17% of the exam. Turning a business problem into an architecture: end-to-end design with feedback loops, choosing between workflow, agentic and augmented patterns, multi-agent orchestration, decomposition, and tying the result to business value.
6 lessons · 1 hr 45 min
From problem to architecture
- Turning a business problem into a Claude solution
- Designing the system end to end
- Choosing between augmented, workflow and agentic
Orchestration and value
- When multi-agent is worth its coordination cost
- Decomposing a problem too large for one call
- Aligning the design to business value
Module 3
Free accountDomain 2: Models, prompting and context engineering
13% of the exam. Selecting a model on real tradeoffs, designing system prompts and guardrails, applying prompting techniques deliberately, engineering the context window, and reusing prompts through caching and Skills.
5 lessons · 1 hr 24 min
Choosing and prompting a model
- Selecting a model you can defend
- System prompts, templates and guardrails that hold
- Choosing a prompting technique under constraint
Context and reuse
- Engineering the context window as a budget
- Reuse: caching, modular prompts and Skills
Module 4
Free accountDomain 3: Integration
The largest domain at 19%. Tool and agent configuration, authentication and authorisation, accuracy against latency, observability at scale, RAG pipeline design, retrieval strategy, connection protocols, and progressive discovery against monolithic context.
7 lessons · 2 hr 3 min
The integration surface
- Evaluating a tool configuration for capability bloat
- Authentication, authorisation and the agent identity gap
- Trading accuracy against latency, and defending the choice
- Observability for agent systems at scale
Retrieval and protocols
- Designing a RAG pipeline: chunking and indexing
- Matching a retrieval strategy to data shape and query pattern
- Choosing a connection protocol and a discovery strategy
Module 5
Free accountDomain 4: Evaluation, testing and optimisation
16% of the exam. Defining metrics that mean something, building evaluation datasets and mixed-method test frameworks, running A/B tests, diagnosing failures to the right cause, and optimising the cost, latency and quality triangle.
6 lessons · 1 hr 41 min
Measuring what matters
- Metrics that survive contact with a stakeholder
- Evaluation sets and mixed-method test frameworks
- A/B testing and the iteration loop
Diagnosing and tuning
- Diagnosing a failure to the right cause
- Optimising cost, latency and quality together
- Monitoring and observability for a running system
Module 6
Free accountDomain 5: Governance, safety and risk management
14% of the exam. Guardrails and safety controls, the real failure modes of language model systems, human-in-the-loop validation, regulated environments, and the ethical considerations an architect is expected to raise.
5 lessons · 1 hr 26 min
Controls and failure modes
- Guardrails that hold when the model does not
- How these systems actually fail
- Designing the human into the system
Compliance and ethics
- Designing for a regulated environment without overclaiming
- Bias, fairness and transparency as design decisions
Module 7
Free accountDomain 6: Stakeholder communication and lifecycle
14% of the exam, and the domain engineers most often underestimate. Structured discovery, communicating a tradeoff to someone who wanted the other option, managing expectations and SLAs, documenting a design, and carrying a system through its lifecycle.
5 lessons · 1 hr 24 min
Discovery and communication
- Discovery that produces a problem statement
- Explaining a tradeoff to the person who wanted both
Through the lifecycle
- Committing to a service level on a probabilistic system
- Documentation the receiving team can operate from
- Carrying a system through its lifecycle
Module 8
Free accountDomain 7: Developer productivity and operational enablement
The smallest domain at 7%. Configuring Claude tooling for a team rather than an individual, improving developer workflow with AI assistance, and supporting debugging and operational resolution.
3 lessons · 47 min
Enabling a team
- Configuring Claude tooling for a team, not a person
- Where AI assistance actually improves a developer workflow
- Debugging and operational issue resolution with AI assistance
Where this course comes from
Independent study guide. Structure follows the publicly published CCAR-P exam blueprint; all explanations, examples and practice questions are our own.
Start with the orientation module
It reads with no account at all, and it sets out what the exam measures, how to prepare for it, and what this guide is and is not sure about.
3 of 40 lessons are open to read without an account, and a free account opens the other 37. There is no paid plan, no tier and nothing to buy.