Claude Certified Architect — Professional (CCAR-P): complete study guide
The Professional exam is not a harder version of Foundations with the same syllabus. It adds three domains that do not exist at Foundations level — governance and risk, stakeholder communication and lifecycle, and developer enablement — which together make up 35% of the paper. The shift is from designing a system that works to defending one that an organisation can measure, govern, explain and still be running in two years.
What separates Professional from Foundations
The Foundations exam asks whether your design works. The Professional exam asks whether you can prove it works, govern it, explain it to the people paying for it, and still be running it in two years. That shift is not rhetorical — it is visible directly in the blueprint, which adds three domains that do not exist at Foundations level and which together make up 35% of the paper: governance and risk, stakeholder communication and lifecycle, and developer enablement.
The audience is mid- to senior-level: solution architects, AI and ML engineers, technical leads and senior engineers, frequently working in regulated sectors. Where Foundations assumes around six months with Claude, Professional assumes years of architecture or platform engineering behind you.
Format — and the change that catches people out
| Scored questions | 63 | Time | 120 minutes |
|---|---|---|---|
| Pass mark | 720 on a 100–1,000 scale | Format | Multiple choice and multiple response |
| Domains | 7 | Validity | 12 months |
| Prerequisites | None — you may sit it without CCAR-F | Delivery | Pearson VUE, online-proctored or test centre |
The Professional exam uses standalone items rather than the scenario-anchored format of Foundations. On CCAR-F, four of six long scenarios frame every question and you can build a mental model of the system once and reuse it. Here each question stands alone, with its own context, and you get roughly 1.9 minutes per item. There is no shared setup to amortise, so the skill being tested is reading a fresh situation quickly and identifying the one constraint that decides the answer.
No prerequisite exam is required, though sitting Foundations first is sensible: the Professional paper assumes the Foundations material rather than repeating it.
Domain weights
| # | Domain | Weight | Roughly | The question behind it |
|---|---|---|---|---|
| 3 | Integration | 19% | ~12 questions | How does this connect to everything else the business runs? |
| 1 | Solution Design and Architecture | 17% | ~11 questions | Is this the right shape for the problem? |
| 4 | Evaluation, Testing and Optimisation | 16% | ~10 questions | How do you know it works, and how do you know when it stops? |
| 5 | Governance, Safety and Risk | 14% | ~9 questions | Will this survive an audit? |
| 6 | Stakeholder Communication and Lifecycle | 14% | ~9 questions | Can you defend it, document it and hand it over? |
| 2 | Claude Models, Prompting and Context Engineering | 13% | ~8 questions | Which model, which context strategy, at what scale? |
| 7 | Developer Productivity and Operational Enablement | 7% | ~4 questions | Can a team actually build and run this? |
Integration is the single heaviest domain at 19% — more than the pure architecture domain. That tells you where the exam thinks the difficulty in real systems lies: not in choosing a pattern, but in connecting it to identity systems, data stores, observability and everything else an enterprise already operates.
The three domains absent from Foundations — governance, stakeholder communication and developer enablement — total 35%. If you are coming from CCAR-F, that is where your preparation gap is, and it is not a gap you close by reading API documentation.
How the questions are built
Expect a short professional situation and four defensible-looking options. Because the items are standalone, the deciding constraint is always stated in the stem, and it is usually one of these:
- A regulatory or contractual requirement — data residency, auditability, a retention obligation, a sector rule.
- A measurable service target — a latency budget, an accuracy floor, a cost ceiling, an availability commitment.
- An organisational reality — who owns the system, who reviews the output, what the team can realistically operate, what a stakeholder was promised.
Find that constraint before you evaluate the options. Two answers will usually be technically sound, and the one that satisfies the stated constraint is the one that scores.
Every page in this guide
- The mental model — Foundations asks whether the design works. Professional asks whether you can prove it, govern it, explain it and still be running it in two years.
- Domain 1 — Solution Design and Architecture — Translating a business problem into a Claude architecture: choosing the shape, naming every component, defining failure behaviour, and defending the design on unit economics.
- Domain 2 — Claude Models, Prompting and Context Engineering — Model tier allocation across a system, context engineering at scale, caching economics, and treating prompts and model versions as governed artefacts.
- Domain 3 — Integration — The heaviest CCAR-P domain: integration mechanisms, identity and authorisation, RAG pipeline design, citations, the accuracy-latency-cost trade and observability.
- Domain 4 — Evaluation, Testing and Optimisation — Eval suites as a deliverable, LLM-as-judge validation, shadow and A/B testing, production monitoring signals, and isolating a failure to the right layer.
- Domain 5 — Governance, Safety and Risk — Named LLM failure modes, guardrails as layers, designing human review that actually happens, the regulatory landscape and building an audit trail.
- Domain 6 — Stakeholder Communication and Lifecycle — Discovery and requirements, communicating trade-offs, what can honestly go in an SLA, the documents that matter, and planning the lifecycle and handover.
- Domain 7 — Developer Productivity and Operational Enablement — Standardising a team’s Claude Code setup, reproducible automation, bounded autonomy and making a system operable by people who did not build it.
- The 14 anti-patterns — Fourteen distractors that recur across the CCAR-P question bank, each with why it fails and what the exam wants instead.
- Cheat sheet — Domain weights, the defensibility test, guardrail layers, the RAG pipeline and its failure modes, compliance references and what can go in an SLA.
- Questions & answers — Domain 1 — CCAR-P practice questions on solution design and architecture, each with the correct answer and a full explanation of why the distractors fail.
- Questions & answers — Domain 2 — CCAR-P practice questions on claude models, prompting and context engineering, each with the correct answer and a full explanation of why the distractors fail.
- Questions & answers — Domain 3 — CCAR-P practice questions on integration, each with the correct answer and a full explanation of why the distractors fail.
- Questions & answers — Domain 4 — CCAR-P practice questions on evaluation, testing and optimisation, each with the correct answer and a full explanation of why the distractors fail.
- Questions & answers — Domain 5 — CCAR-P practice questions on governance, safety and risk, each with the correct answer and a full explanation of why the distractors fail.
- Questions & answers — Domains 6–7 — CCAR-P practice questions on discovery, communicating trade-offs, SLAs, handover and team enablement, with answers and full explanations.
- Facts to memorise — Thirty recall facts covering guardrail layering, the RAG pipeline, citations, compliance attestations, monitoring signals and SLA limits.
- Six-week study plan — A week-by-week CCAR-P plan that pairs each domain with something you build, measure or write, because this exam rewards experience over reading.
- Official resources — Contextual retrieval, citations, cloud deployment surfaces, eval design, jailbreak mitigation and the Trust Portal, mapped to the seven domains.
This is an independent study guide, not an official Anthropic publication. Exam blueprints, pricing and product details change; confirm the current domain weights and format against the official exam guide before you sit.
Next: The mental model →
Explore Other Certification Study Guides
Switch to another exam guide to prepare for your next Claude certification:
Claude Certified Associate — Foundations
Entry-level certification for non-coders: prompting, output evaluation, model selection, workflow integration, configuration, governance & responsible use.
Claude Certified Architect — Foundations
Core architectural principles: agentic loops & orchestration, tool design & MCP, Claude Code workflows, prompt engineering, and context reliability.
Claude Certified Developer — Foundations
Build software on Claude: API integration, model selection, tools & MCP servers, agents vs workflows, security, safety & debugging.