Anthropic · CCAR-P · Professional

Claude Certified Architect – Professional (CCAR-P)

Exam update: Anthropic has announced that an updated exam guide will be available from 17 November 2026 and that the exam will include updated content from 7 December 2026. Taking the exam on or after 7 December 2026? Study from the updated guide. Taking it before then? Study from the current guide (v1.0). We'll update this page once the new guide is out.

Mid- to senior-level solution architects, AI/ML engineers, technical leads and senior engineers who design, build and deliver production Claude solutions at enterprise scale and lead architectural/stakeholder decisions. Not for entry-level developers, casual users, or non-technical/prompt-only roles.

60 free practice questions · full timed mock exam · study plan

Exam facts

Checked 2026-11-15 against the official exam guide (Version 1.0, effective July 2026 (initial publication July 2026)). Official exam guide · Official certification FAQ

Retakes, rescheduling and more

What the exam covers

D1 Solution Design & Architecture (17% of the exam)

D2 Claude Models, Prompting & Context Engineering (13% of the exam)

D3 Integration (19% of the exam)

D4 Evaluation, Testing & Optimization (16% of the exam)

D5 Governance, Safety & Risk Management (14% of the exam)

D6 Stakeholder Communication & Lifecycle Management (14% of the exam)

D7 Developer Productivity & Operational Enablement (7% of the exam)

Sample practice questions

D1

An insurer wants Claude to process supplier invoices. Every invoice goes through the same steps: extract fields, check them against the purchase order, then route to the right approver. Volume is high and auditors want predictable behaviour. Which architecture is the BEST fit?

  • An autonomous agent given all ERP tools that plans how to handle each invoice on its own
  • A coordinator agent that delegates extraction, matching and routing to three specialist subagents
  • A prompt chain with code-based validation gates between extraction, matching and routing
  • One large prompt that extracts, matches and routes each invoice in a single model call

D1

A consulting firm wants Claude to investigate competitor markets for clients. Some questions need two searches; others need dozens of follow-ups that depend on what earlier searches reveal. Which property of this problem MOST justifies an agentic design rather than a fixed workflow?

  • The steps can't be predicted up front and depend on what earlier searches find
  • The firm needs to process many requests per day at a predictable cost per request
  • Clients expect the output to look more sophisticated than a standard chatbot reply
  • Each research request must follow the same review steps defined by firm policy

D1

An HR team wants an assistant that answers employee questions from regional policy handbooks that total several thousand pages and change weekly. An engineer proposes a five-agent system with a planner, retriever, writer, checker and editor. No evals exist yet. What should the architect recommend FIRST?

  • Build the five-agent design now but cap each agent at three turns to keep spending under control
  • Start with one Claude call augmented with retrieval; add steps only if evals show gaps
  • Fine-tune a separate model on the handbook so that retrieval is not needed to answer questions
  • Place every handbook in full in each request and ask Claude to answer directly from them

Answer these and more, with every option explained