What the Claude Certified Architect – Professional exam actually asks.

• 63 questions in 120 minutes — about 1 minute 55 seconds each. The guide's published sample items are standalone; MockRoom's practice items use that same independent-question format.
• It sits at the Professional level of the Claude certification path, and tests solution design, model selection, integration architecture, evaluation, governance and lifecycle management.
• Your result is reported as a pass or fail with a scaled score from 100 to 1,000, and the report also gives the percentage you answered correctly within each content domain, as a guide to where you stand.
• 720 out of 1000 to pass, on a scaled score. The exam is criterion-referenced: you are measured against a fixed performance standard rather than against other candidates, so the raw number of correct answers you need is not fixed or published.
• Question formats: multiple choice and multiple response. A multiple-response item tells you how many to select and expects exactly that many.
• Answer everything. A blank scores nothing, so flag the long ones, bank the quick ones, and return before time runs out.
• $175 USD to sit, and the fee applies to each attempt.
• Proctored delivery: online proctored and/or at a test centre, according to the certification programme's policy. Registration and scheduling run through the Anthropic Partner Academy and Pearson VUE.
• Validity and renewal: 12 months from the date the credential is awarded. Complete the free, non-proctored renewal assessment on time to renew; after a lapse, or when a significant exam change requires it, regain certified status by retaking the full exam at the applicable fee.
• Retake policy: the waiting period increases with each failed attempt — 14 days after the first, 30 days after the second, 90 days after the third. You may take the exam up to four times within a rolling twelve-month period.

Two hours for sixty-three items rewards precise reading. Decide early how long an item may take before you flag it and move on.

• Domain 1 — Solution Design & Architecture (17%). Translating business problems into Claude-based solutions; end-to-end architectures across input, processing, output and feedback loops; choosing between workflow, agentic and augmented LLM patterns; multi-agent systems and orchestration strategies; decomposition techniques for complex problem solving; and aligning a design to business value pillars.
• Domain 2 — Claude Models, Prompting & Context Engineering (13%). Selecting Claude models against trade-offs; designing system prompts, templates and guardrails; zero-shot, few-shot and chain-of-thought prompting; optimising context windows and managing token usage; and prompt reuse through caching, modular prompts and Skills.
• Domain 3 — Integration (19%). Tool and agent configuration and the problem of capability bloat; authentication and authorisation requirements; accuracy-latency trade-offs; observability and monitoring at scale; RAG pipeline design with chunking and indexing; retrieval strategies matched to data shape and query pattern; integration mechanisms including MCP, API/CLI and agent-to-agent; and progressive discovery against monolithic context.
• Domain 4 — Evaluation, Testing & Optimization (16%). Defining evaluation metrics across accuracy, latency, cost, safety and security; designing evaluation datasets and test frameworks; A/B testing and iterative improvement; diagnosing prompt failure, hallucination and model mismatch; optimising tokens, latency and cost-performance; and monitoring with logging and observability tools.
• Domain 5 — Governance, Safety & Risk Management (14%). Implementing guardrails and safety controls; identifying risks, limitations and failure modes; human-in-the-loop validation strategies; compliance with regulations such as GDPR, HIPAA and FedRAMP; and ethical AI considerations including bias, fairness and transparency.
• Domain 6 — Stakeholder Communication & Lifecycle Management (14%). Structured discovery and requirement gathering; communicating architectural decisions and trade-offs; managing stakeholder feedback loops and expectation alignment, including SLAs; documenting architectures and providing implementation guidance; and the lifecycle phases from discovery through design and handoff to monitoring and iteration.
• Domain 7 — Developer Productivity & Operational Enablement (7%). Configuring Claude tools and environments for teams, including Claude Code; improving developer workflows with AI-assisted tooling; and supporting debugging and operational issue resolution.

Domain 3 (Integration) is the largest at 19%, and the top three domains — Integration, Evaluation, and Solution Design & Architecture — together make up 52%. The smallest domain, Developer Productivity & Operational Enablement, is worth about 4 questions per form; it is a low-effort mark if you know the mechanics precisely.

Expect a short item — a design choice, a trade-off, a governance call — followed by a question about the best approach, the right mechanism, or the correct prioritisation. More than one option may work in isolation; the exam asks for the one that satisfies every stated constraint. The guide's intended audience is mid- to senior-level professionals who design, build and deliver production-grade AI solutions - chiefly solution architects, AI/ML engineers, technical leads and senior software engineers. Anthropic publishes the exact task statements in the exam guide, so confirm them there before you sit the paper.

Read the question stem first, then the scenario, then the options. Qualifiers to look for:

• “Deterministic” → Prefer a programmatic gate or a schema constraint over asking the model to remember.
• “At scale” or “within budget” → Ask where the tokens or the latency actually go before reaching for a bigger model.
• “Auditable” or “traceable” → Look for the enforcement point outside the prompt, not inside it.
• “From fetched content” or “user-supplied” → Treat the text as untrusted and look for the validation layer.
• “Without retraining” → Favour RAG, prompt or tool changes over model customisation.

This exam tests design decisions you own, so the questions reward knowing mechanics exactly: which retrieval mode matches an exact identifier, which hook fires before a tool runs, which cache breakpoint invalidates a prefix, which evaluation metric catches regression. Vague familiarity is what the wrong options are written against.

Three habits that work:

• Build a RAG pipeline end to end. Retrieval, reranking, generation, evaluation, and observability touch most of the domains and make the questions concrete.
• Read the Anthropic documentation alongside the guide. The Messages API, the Agent SDK, MCP, and the evaluation cookbook each have their own reference; the guide names the topics but the documentation settles the detail.
• Say why each wrong option fails. On an exam about constraints, learning to name the violated constraint is the skill being tested — and it is also how wrong options are constructed.

MockRoom's CCAR-P study room has five distinct 63-question mock exams — 315 original questions in total, written against the published exam guide and weighted across its seven domains, each with an explanation for every option. It also has 302 active recall flashcards for spaced review. Nothing here is taken from an exam dump or from the guide's own sample questions, and none of it is real examination content.

Open the CCAR-P study room