Anthropic · CCAR-P · Professional
Claude Certified Architect – Professional
Architecting Claude at enterprise scale: solution design, models and context engineering, integration, evaluation, governance and safety, stakeholders and lifecycle. Currently for Claude Partner Network staff.
Official exam page ↗What the exam covers
- Solution Design & Architecture17%
- Claude Models, Prompting & Context Engineering13%
- Integration19%
- Evaluation, Testing & Optimization16%
- Governance, Safety & Risk Management14%
- Stakeholder Communication & Lifecycle Management14%
- Developer Productivity & Operational Enablement7%
Weights are the vendor's published exam guide.
Step 1 · Assess
Start with a 20-question diagnostic
Free, no account. Drawn across every domain in the exam’s own weights, so the score means what the real thing will. Treat it like the exam and don’t look anything up — the map is only useful if it’s true.
About the Claude Certified Architect – Professional (CCAR-P) exam
The Claude Certified Architect – Professional is Anthropic's senior certification for architects who take Claude into production across an organisation. It goes beyond design choices to the whole life of a solution: integration with enterprise systems, evaluation, governance and safety, and bringing stakeholders with you.
The exam guide, effective July 2026, has seven domains. Integration is the largest, at 19%, followed by solution design and architecture, evaluation, testing and optimisation, governance, safety and risk management, stakeholder communication and lifecycle management, Claude models, prompting and context engineering, and developer productivity and operational enablement. Questions are scenarios with competing constraints, where more than one design could work and one is best.
At the time of writing the certification is open only to staff of organisations in the Claude Partner Network. Anthropic publishes free prep courses; this practice is ours, written against the public exam guide.
The exam at a glance
- Questions
- 63 questions, multiple choice and multiple response
- Time
- 120 minutes
- Pass mark
- 720 on a scale of 100 to 1,000
- Exam fee
- About US$175
- Valid for
- 12 months; renew before it lapses with Anthropic's free renewal assessment
- Where you sit it
- Pearson VUE, online proctored or at a test centre
- Prerequisites
- None required, but you must work for a Claude Partner Network organisation to register
As published in the vendor’s exam guide at the time of writing. Fees vary by country and change; the vendor’s page is the source of truth.
What is on the exam, domain by domain
Solution Design & Architecture
Designing Claude solutions for real constraints: scale, cost, latency, reliability and data requirements.
17% of the paper
Claude Models, Prompting & Context Engineering
Choosing models and capabilities, and engineering prompts and context for production use.
13% of the paper
Integration
The largest domain. Connecting Claude to enterprise systems, platforms and data, including the cloud platforms that offer Claude.
19% of the paper
Evaluation, Testing & Optimization
Defining success criteria, building evaluations, testing changes, and optimising quality, latency and cost.
16% of the paper
Governance, Safety & Risk Management
Policy, safety controls, risk assessment and oversight for Claude in an organisation.
14% of the paper
Stakeholder Communication & Lifecycle Management
Explaining trade-offs to stakeholders, managing change, and running a solution through its life.
14% of the paper
Developer Productivity & Operational Enablement
Helping engineering teams work effectively with Claude, including Claude Code.
7% of the paper
How to prepare
- Read the exam guide's task statements and map each to a production decision you have made or seen. This exam rewards experience.
- Know how Claude is offered and integrated across Anthropic's API and the major cloud platforms, and what changes between them for security and data residency.
- Learn how to define success criteria and build an evaluation before you change a prompt or a model; evaluation is 16% of the paper.
- Practise scenario questions by naming the deciding constraint first, then eliminating designs that break it.
- Take Anthropic's free prep course, then practise papers here until you clear your target twice running.
Sample questions
Three questions in the exam’s style, written by Square 1 against the published domains. They are ours, not the vendor's. The answer and the reasoning are under each one; the practice papers on this page draw from a bank written the same way, with an explanation on every question.
Question 1 · Evaluation, Testing & Optimization
A team wants to switch its production assistant to a cheaper model to cut costs. What should happen before the switch?
- Run the new model against an evaluation set built from real tasks and compare the results
- Switch in production and watch for complaints from users over the first week
- Ask the new model whether it can handle the current workload before switching
- Increase the prompt length so the cheaper model receives more guidance
Answer: A. A model change should be judged against defined success criteria on representative tasks before it reaches users. Waiting for complaints measures failure after the fact, and a model's own claim about its ability is not evidence.
Question 2 · Integration
A regulated client requires that its data stays within its existing cloud provider's controls and billing. How is this most directly met?
- Use Claude through that cloud provider's platform where it is offered, inside the client's account
- Send the data to the public Claude API and ask Anthropic to delete it after each request
- Encrypt the data before each request so it can be sent to any endpoint safely
- Run the model on the client's own laptops so no data leaves the building
Answer: A. Claude is offered through major cloud platforms, so a client can keep usage within its existing provider's controls, security and billing. Encrypting a prompt does not work because the model needs to read it, and Claude is not run on client laptops.
Question 3 · Stakeholder Communication & Lifecycle Management
A pilot shows strong results but senior stakeholders worry about errors in customer-facing replies. What is the best next step?
- Propose a staged rollout with human review of replies and agreed quality measures to expand on
- Launch to all customers at once because the pilot results were strong
- Pause the project until the model can guarantee it will never make an error
- Hide the use of Claude from customers so any errors do not affect trust
Answer: A. A staged rollout with human review and agreed measures answers the concern with evidence and control. No model can guarantee zero errors, and hiding AI use from customers creates a trust and governance problem of its own.
Questions people ask before they book
- Who can take the Claude Certified Architect – Professional exam?
- At the time of writing, only staff of organisations in the Claude Partner Network, registering with a partner email address. Anthropic has not announced a date for opening it to everyone.
- Do I need the Foundations certification first?
- No certification is formally required, but the Professional exam assumes the knowledge the Foundations exams test and adds production experience on top.
- What makes the questions hard?
- Most are scenarios where several designs would work and one is best under the stated constraints. Name the constraint that decides it before looking at the options.
- How long is the certification valid?
- Twelve months, renewable before it lapses with a free renewal assessment; after a lapse you sit the full exam again.
- Are these the real exam questions?
- No. Every question here is ours, written against Anthropic's public exam guide. Anthropic's exam content is confidential, and practising on leaked questions would break its exam policy.
