CCAR-P
Free Claude Certified Architect - Professional Practice Test
12 exam-style questions with full explanations — no sign-up. Score yourself, then close your gaps with the full course.
Exam fee ~$2507 exam domainsLevel Advanced2 timed practice tests in the course
✅ Free practice — no sign-up📝 Real exam-style questions💡 Detailed explanations💸 30-day money-back via Udemy
An internal MCP server lets Claude query a CRM system on behalf of signed-in employees. Different employees should only see records they are authorized to view in the CRM, and the security team has rejected any design using a single shared privileged credential. Which authentication approach BEST meets these requirements?
On-behalf-of delegation preserves the end user's existing permission scope in the downstream CRM, giving per-user least privilege without the MCP server needing elevated standing access.
A SaaS company is building an internal engineering assistant that answers questions about system architecture by searching docs, querying a metrics API, and occasionally writing a small script to analyze logs. Each request requires 2-4 sequential tool calls, and the exact sequence of tools needed cannot be predetermined because it depends on what earlier tool calls return. Which architecture pattern BEST fits this requirement?
When the sequence and number of tool calls cannot be known ahead of time and depends on intermediate results, Anthropic's guidance recommends an agent loop where the model autonomously plans, calls tools, observes results, and decides the next action until the task is complete.
A team built a customer support agent and needs to evaluate response quality at scale before each deployment. They need fast, automated scoring across hundreds of test cases without requiring exact-match ground truth for open-ended replies. Which evaluation approach BEST fits this requirement?
LLM-as-judge with a well-defined rubric enables scalable, automated, nuanced quality assessment that correlates well with human judgment on open-ended generative tasks, which exact-match or n-gram metrics cannot handle.
A healthcare company wants to build a clinical documentation assistant using the Claude API that will process Protected Health Information (PHI). Which action is REQUIRED before sending PHI to the API to remain HIPAA compliant?
HIPAA requires a signed BAA with any vendor (business associate) that creates, receives, maintains, or transmits PHI on behalf of a covered entity. Anthropic offers BAAs to qualifying enterprise accounts, and this legal agreement is a prerequisite, not a technical setting.
A production application calls the Claude API using the alias "claude-3-5-sonnet-latest". After Anthropic updates which snapshot the alias points to, the architecture team notices subtle changes in response tone and formatting that break a downstream parser. Which change-management practice would BEST have prevented this incident?
Dated model snapshots are immutable, so pinning to one guarantees the model behavior never changes underneath the application. Upgrades then become a deliberate, tested migration event rather than a silent runtime change — this is the core lifecycle-management practice for production systems using Claude.
An architect is building a pipeline that classifies 2 million customer support tickets per day into one of 12 categories. Latency must stay under 300ms per request and the solution must minimize per-token cost while maintaining acceptable accuracy. Which model should be selected for this workload?
Haiku is purpose-built for high-volume, latency-sensitive, lower-complexity tasks such as ticket classification, delivering the best cost/latency profile for this use case.
A platform engineering team is rolling out Claude Code to 500 engineers across hundreds of repositories. They want every engineer, regardless of which repo they're working in, to automatically get consistent project-specific context (build commands, test commands, coding conventions) without retyping it in every prompt or relying on tribal knowledge. Which approach BEST meets this requirement?
A repository-root CLAUDE.md, committed to version control, is automatically loaded by Claude Code for anyone working in that repo, giving every engineer consistent, up-to-date project context with zero repeated effort and a single source of truth that evolves with the codebase.
A legacy on-premises database only exposes a JDBC interface, with no REST, GraphQL, or SOAP endpoint available. Claude needs read-only access to a few specific views in this database for an internal reporting agent. Which integration approach is architecturally correct?
Claude does not speak raw database protocols; the correct pattern is an adapter service (commonly an MCP server) that performs the protocol translation and enforces scope, exposing only the approved, parameterized operations as tools.
A company must classify 2 million short support tickets per day into one of 12 categories, with a strict budget and a latency requirement under 500ms per ticket. Accuracy only needs to be 'good enough' to beat the existing keyword-based system. Which model choice is MOST appropriate?
High-volume, latency-sensitive, low-complexity classification is the canonical use case for the fastest, cheapest model tier; Haiku-class models meet the 'good enough' quality bar at the throughput and cost this workload demands.
A RAG application sends a large static system prompt (5,000 tokens) and a large reused reference document (20,000 tokens) with every request, followed by a short unique user question. To minimize token cost across thousands of repeated requests using prompt caching, where should cache_control breakpoints be placed?
cache_control markers should be placed immediately after stable, reused content blocks (system prompt, document) so that identical prefixes hit the cache across requests, while the variable user question is excluded so it doesn't invalidate caching.
A European financial institution must ensure that all customer data sent to Claude remains within EU infrastructure boundaries to satisfy GDPR and national data-residency regulations. Which deployment approach BEST satisfies this requirement?
Amazon Bedrock and Google Vertex AI both offer regional deployments of Claude that keep inference processing and associated data within the selected region, which is the architecturally supported way to meet EU data-residency requirements.
A business stakeholder asks why the architecture team routes simple intent-classification requests to Claude Haiku while routing multi-step contract analysis to Claude Opus, when both could technically call the same model. What is the BEST justification to give in this conversation?
A core architect responsibility is defending model-selection decisions in terms stakeholders care about: cost, latency, and accuracy tradeoffs. Tiered routing by task complexity is the standard cost/performance optimization pattern and gives a concrete, defensible rationale.
Ready for the real thing?
The full course has two full-length practice tests, video lessons for every exam domain, hands-on labs and detailed answer explanations.
Start my full course on Udemy →CCAR-P exam — quick answers
How much does the CCAR-P exam cost?
The exam fee is approximately $250 and varies by region — confirm current pricing with the certification vendor before you book.
What topics are on the exam?
It covers 7 domains: Integration (19%), Solution Design and Architecture (17%), Evaluation, Testing, and Optimization (16%), Governance, Safety, and Risk Management (14%), Stakeholder Communication and Lifecycle Management (14%), Claude Models, Prompting, and Context Engineering (13%), Developer Productivity and Operational Enablement (7%). The full course has a dedicated chapter, lab and practice-test coverage for each.
Is this practice test really free?
Yes — all questions on this page are free with explanations and no sign-up. The paid Udemy course adds two full-length timed exams, video lessons and hands-on labs.
Will this prepare me for the real exam?
The questions mirror the real exam's style and are mapped to the official domains. This is exam-focused preparation — combine the free test with the full course's timed simulations to gauge your readiness.