Free lesson · Claude Certified Architect - Professional (CCAR-P)

What CCAR-P Tests and How It Differs from Foundations

Easy15 min readorientationccar-pexam-overviewpositioning

The Exam in One Sentence

CCAR-P asks whether you can take a business problem, turn it into a Claude-based system, put that system into production, and keep it alive.

That last part is where it separates from every other Claude certification. Building something that works is assumed. What the exam actually measures is everything around the building - the decisions before it, the defence of those decisions, and the ownership after launch.

The Facts

Warning: Verified 3 September 2026 against publicly available material describing Exam Guide v1.0 (effective July 2026). Anthropic can change weights, objectives, and mechanics without notice. Confirm against the current official exam guide before you sit.

No Prerequisites - Read That Carefully

You do not need to pass Foundations first. You do not need any prior certification. The credential is awarded on exam performance alone.

That is a genuine door, and it is worth being clear about what it does and does not mean.

It means you can sit Professional directly if your experience supports it.

It does not mean Professional is a natural next step after Foundations. The two exams overlap less than the shared name suggests, and roughly half of what Professional tests does not appear in Foundations at all.

Important: Do not assume passing Foundations puts you most of the way to Professional. Three of the seven Professional domains - governance, stakeholder communication and lifecycle, and much of integration - have no Foundations equivalent. Together those are close to half the exam.

The Recommended Background

Anthropic's guide recommends two things.

Three or more years of systems architecture or platform engineering. Not three years of AI work - three years of architecture. The exam draws on ordinary senior engineering judgment: service levels, failure domains, capacity, compliance, stakeholder management. Claude is the subject; architecture is the discipline.

Six or more months of hands-on production work with Claude or comparable LLM systems.

Both are recommended, not required. Nobody checks. But they describe who the questions were written for, and reading the exam without that background means meeting concepts as new material rather than as familiar problems in a new setting.

Tip: If you have the Claude experience and not the architecture experience, the domains to watch are stakeholder communication and lifecycle (14%) and governance (14%). Those draw hardest on general senior practice and least on anything Claude-specific. If you have the architecture experience and not the Claude depth, the models and context domain is where to concentrate.

Foundations Versus Professional

The clearest way to see the difference is what each exam assumes you start with.

graph TD
    A["Foundations starts with<br/>a defined task"] --> B["Asks: can you<br/>build it well?"]
    C["Professional starts with<br/>a business problem"] --> D["Asks: should you build it,<br/>and can you own it?"]

This diagram contrasts the two exams. Foundations starts with a defined task and asks whether you can build it well. Professional starts with a business problem and asks whether you should build it, and whether you can own it.

Foundations hands you a task. Design the agent, choose the tools, manage the context, handle the failures. Every question sits inside a system somebody has already decided to build.

Professional hands you a situation. Somebody wants a thing. Should it be built? With what? At what cost? Who signs off? What happens when it goes wrong at three in the morning? Who owns it in eighteen months?

Where the Weight Moved

The domain weights tell the story more precisely than any description.

Two things stand out.

Claude Code drops from 20% to 7%. In Foundations it is the domain candidates most reliably under-prepare relative to its weight. In Professional it is the smallest domain on the exam. If you came from Foundations, this is where your preparation is most over-invested.

Four subject areas appear from nothing. Retrieval, compliance, stakeholder work, and lifecycle ownership. None of them is optional and together they are a large share of the paper.

Important: The single most useful thing a Foundations-holder can know is that their strongest domain is now worth 7%. Time spent revising Claude Code for Professional is time not spent on integration, governance, or stakeholder communication - which are 47% between them.

What "Professional" Means Here

Certification tiers usually mean harder questions on the same material. That is not what is happening here.

CCAR-P is described as a breadth exam. It expects competence across all seven areas rather than depth in one or two. A candidate who is excellent at solution design and weak at governance does not pass on the strength of the first.

graph LR
    A["Foundations<br/>Depth in fewer areas"] --> B["Scenario clusters<br/>4 of 6 archetypes"]
    C["Professional<br/>Breadth across seven"] --> D["Standalone items<br/>38 objectives"]

This diagram contrasts the shape of the two exams. Foundations tests depth in fewer areas through scenario clusters covering four of six archetypes. Professional tests breadth across seven domains through standalone items spanning thirty-eight objectives.

That breadth expectation is why the objective count rose from 29 to 38 while the item count rose only from 60 to 63. Fewer items per objective, spread wider.

Warning: A weak domain is harder to hide on a breadth exam. With 38 objectives and 63 items, most objectives get one or two questions - so a domain you skipped is not a handful of items you might get lucky on, it is a systematic loss spread across the paper.

The Question Style

Two properties are worth knowing before you see one.

Most items are judgment, not recall. You are rarely asked what something is. You are asked which of four defensible approaches fits a stated situation.

Two options are usually both reasonable. The constraint stated in the question is what decides between them - a mentioned budget, a compliance requirement, a team size, an existing system. Reading past that constraint is the most common avoidable error.

Example: The shape of a typical item:

A financial services client needs document summarisation across 40,000 archived contracts. The data cannot leave their region. They have an existing search infrastructure. Which approach fits?

Four options follow, and at least two would be sound architecture in general. The residency requirement and the existing infrastructure are what eliminate the others.

A candidate who reads for the technical question and skims the constraints will find two options equally attractive and pick on instinct.

Tip: Train yourself to note the constraints before reading the options. Region, budget, existing systems, team size, regulatory context, latency requirement. Those sentences are not scene-setting - they are the answer key.

Where This Sits Among the Claude Certifications

Anthropic runs four certifications, ranging roughly $99 to

CCAR-P is the senior architect credential of the set. It is the one that assumes you will be in the room when a director asks why the system costs what it costs.

The Four Domains That Are New Territory

Worth naming them individually, because a Foundations-holder will meet each as unfamiliar ground.

Retrieval and RAG. Chunking, indexing, embedding design, hybrid search, reranking, and - the part most often skipped - measuring whether retrieval is any good. This sits inside the heaviest domain.

Compliance. GDPR, HIPAA, and FedRAMP as they apply to LLM architectures, plus choosing a compliance boundary and handling bias, fairness, and explainability.

Stakeholder communication. Requirements discovery, framing trade-offs for people who do not read architecture diagrams, running a review, and defending a decision to an executive.

Lifecycle ownership. Service levels, error budgets, release gates, decision records, handoff, and deciding when to retire a system.

Tip: If you hold Foundations, these four are where your preparation time belongs. They are unfamiliar rather than harder, and unfamiliar is the more expensive kind of gap on a breadth exam because there is no partial knowledge to fall back on.

What This Course Assumes

That you have architecture experience. We explain Claude-specific mechanisms in full. We assume you already know what an SLO is, why multi-tenancy is hard, and what a compliance boundary means in your industry.

That you want decision procedures, not descriptions. The exam tests choices. Every module ends with a way of deciding, not a summary of what exists.

That you may or may not have taken Foundations. Where content overlaps, we say so and give the short version. Where Professional goes further, we mark it.

Tip: If you hold CCAR-F, read the overlap markers and move quickly through them. If you do not, nothing here depends on having taken it - the material is written to stand alone.

In Short

CCAR-P is a 63-item, 120-minute exam scored 100–1000 with a 720 pass, covering seven domains and 38 objectives at

There are no prerequisites, and Foundations is not a stepping stone - roughly half of Professional has no Foundations equivalent, including retrieval, compliance, stakeholder communication, and lifecycle ownership.

The recommended background is three years of systems architecture plus six months of production Claude work. The architecture experience matters more than the Claude experience for the heaviest new domains.

It is a breadth exam. Claude Code drops from Foundations' 20% to 7%, and a weak domain is difficult to hide across 38 objectives. Most items are judgment rather than recall, and the constraint stated in the question is usually what separates two otherwise defensible options.

Interview: The habit to build from this topic is reading for constraints. Every scenario in this course states conditions that eliminate options - a budget, a region, an existing system, a team size. That reading discipline is worth more marks than any single domain, because it applies to all 63 items.

Exercise: Starter. State the pass mark, item count, and time. Then name the four subject areas Professional covers that Foundations does not.

Intermediate. Rate yourself 1–5 on each of the seven domains before reading further. Keep the numbers - topic 5 uses them.

Advanced. If you hold CCAR-F, list what you would have revised for Professional based on that experience. Compare against the weights. Most Foundations-holders over-invest in Claude Code and under-invest in integration.

Quiz: Quick check: A candidate holds CCAR-F and plans to prepare for CCAR-P by deepening their strongest Foundations areas - agentic architecture and Claude Code. What is wrong with this plan?

A) Nothing - those domains transfer directly and depth always helps B) Claude Code is 7% of Professional and agentic architecture is folded into a 17% domain, while integration, governance, and stakeholder work total 47% and have little or no Foundations coverage C) CCAR-F must be retaken before attempting CCAR-P D) Professional tests the same domains at a higher difficulty

Answer: B - the weights moved substantially. The plan concentrates effort on the smallest domain and one that has been absorbed, while leaving nearly half the exam unaddressed. Option D describes how certification tiers usually work and not how these two relate.

Back to Claude Certified Architect - Professional (CCAR-P)