Optivise AI OptiviseAI
Optivise AI Academy · Certification Prep

Claude Certified Architect
Mock Exam

Scored with the real exam's weighted rules across all five domains. A 60-minute timer, out of 1,000, pass at 720 — get your score, a per-domain breakdown and a full answer review the moment you submit.

← Back to the study guide

Independent study resource — not affiliated with, endorsed by, or sponsored by Anthropic. Questions are original to Optivise AI Academy, based on Anthropic's official documentation and published exam guide.

Sample questions

Five Sample Questions

One question from each exam domain, with the answer and why. All five come from the mock exam's question bank above.

Agentic Architecture & Orchestration · In the canonical Claude agent loop, what is the correct order of the core cycle that repeats until the task is done?+
  1. A. Take action → gather context → verify work
  2. B. Gather context → take action → verify work
  3. C. Verify work → take action → gather context
  4. D. Take action → verify work → gather context

Answer: B. The agent loop is gather context → take action → verify the result → repeat. Acting then verifying is exactly what lets the agent self-correct.

Claude Code Configuration & Workflows · What is the primary role of a CLAUDE.md file in a Claude Code project?+
  1. A. It stores your API key
  2. B. It is project memory / instructions automatically loaded into context to guide Claude in that repo
  3. C. It is a changelog required for deployment
  4. D. It defines the CI pipeline

Answer: B. CLAUDE.md is project memory — conventions, commands, guardrails — pulled into context each session so Claude follows the project's rules.

Prompt Engineering & Structured Output · Why are XML tags (e.g. <document>…</document>) commonly recommended in Claude prompts?+
  1. A. They are required syntax or the API errors
  2. B. They clearly delimit sections so Claude can reliably distinguish instructions, context and data
  3. C. They reduce token cost to zero
  4. D. They switch the model to a different family

Answer: B. Claude responds well to XML-style tags because they unambiguously separate parts of the prompt (context vs instructions vs examples), improving reliability.

Tool Design & MCP Integration · What most improves whether Claude calls a tool correctly?+
  1. A. A short cryptic name with no description
  2. B. A clear, detailed description of what the tool does and when to use it, plus well-typed parameters
  3. C. Making every parameter a free-form string
  4. D. Hiding the parameters from the schema

Answer: B. The model relies on the tool's name, description and JSON Schema to decide when and how to call it; clear descriptions and precise types drive correct usage.

Context Management & Reliability · You send the same large system prompt and document on every request. How do you cut latency and cost?+
  1. A. Use prompt caching to reuse the stable prefix across requests
  2. B. Delete the system prompt
  3. C. Lower the passing score
  4. D. Switch to a smaller context window

Answer: A. Prompt caching stores a stable prefix (system prompt, long docs) so repeated requests skip re-processing it — big latency and cost savings.