Skip to main content
Artificial Intelligence ⭐ Premium ⭐ Featured

Anthropic Claude Certified Architect – Foundations - 400 Questions

By Webmaster Certland ❤️ 0 likes

Practice exam for the Anthropic Claude Certified Architect – Foundations certification. Covers agentic architecture & orchestration, tool design & MCP integration, Claude Code configuration & workflows, prompt engineering & structured output, and context management & reliability.

🔒

Premium Content

This exam is exclusive to Premium users. Upgrade to get unlimited access!

Become Premium

👁️ Free Preview (5 of 400 questions)

1. A company is building an agentic application that calls Claude via the Messages API. After each API response, the application checks the stop_reason field. The response returns stop_reason: "tool_use". What should the application do next?

A Extract all tool_use blocks from the response, execute each tool, and send the tool results back as a new user message to continue the loop.
B Present the assistant's partial response to the end user, since the model has provided a complete answer up to the tool invocation.
C Terminate the agentic loop because the model has reached a decision point and no further API calls are needed.
D Retry the same API call with identical messages, waiting for the model to produce an end_turn response instead.

2. An architect is designing an agentic system where Claude must complete a multi-step data analysis task. The API response returns stop_reason: "end_turn" with a full narrative answer in the content array. What is the correct next action?

A Continue the agentic loop by appending the assistant's response to messages and making another API call to verify the answer is complete.
B Exit the agentic loop and present the assistant's response content to the end user as the final answer.
C Scan the response content for any tool_use blocks that may have been included alongside the text and execute them before presenting the answer.
D Send an empty tool_result message back to Claude to acknowledge receipt of the final answer before closing the session.

3. A company is building a customer support agent using Claude. The agentic loop receives a response where stop_reason is "max_tokens". The response content contains a partially completed troubleshooting procedure. What should the architect design the system to do?

A Present the partial response content to the user as the final answer, since Claude has provided all it can within the token budget.
B Terminate the session immediately and return an error to the user indicating the agent failed to complete the task.
C Recognize that the response is incomplete, handle the truncation gracefully by either increasing max_tokens or prompting Claude to continue, and do not present partial output as a final answer.
D Restart the entire conversation from the beginning with a fresh messages array to avoid sending incomplete context.

4. An architect is reviewing the basic structure of an agentic loop in application code. Which of the following pseudocode patterns correctly represents the Claude agentic loop?

A Call the API once, parse the text response for tool name keywords, execute any tools found by string matching, and repeat until no tool keywords appear in the text.
B Call the API, immediately send tool results back without appending the assistant message, check stop_reason, and break when stop_reason is "end_turn".
C Configure Claude with a loop parameter so that it internally repeats tool calls until it determines the task is complete, then returns a single final response.
D Call the API, check stop_reason; if "end_turn" break and return the response; if "tool_use" extract tool_use blocks, execute all tools, append assistant and tool_result messages, then loop again.

5. A company is developing an agentic pipeline where Claude may call multiple tools in a single response. The API response contains three tool_use blocks in the content array. What is the correct way to handle this situation before the next API call?

A Execute all three tools, collect all three results, and send a single user message containing all three tool_result content blocks before making the next API call.
B Execute only the first tool_use block, send its result back, and wait for Claude to request the remaining tools in subsequent responses.
C Make three separate API calls, each containing one tool_result, to ensure the model processes each result independently before proceeding.
D Send the three tool_use blocks back to Claude without executing them, and allow the model to decide which results are necessary before committing to execution.

Want to test yourself for real?

Create a free account and run our exam simulation engine.

Free No credit card
  • Simulation engine
  • Up to 10 questions per attempt
  • Score & basic stats
Create free account Already have an account? Sign in
Best
Premium 7-day trial
  • All 400 questions
  • Detailed explanations
  • Smart Practice + Focus Mode
⭐ Start 7-day free trial

Information

Questions 400
Time 1h 30min
Difficulty Hard
Minimum Score 72.00%
▶ Try 5 questions — no account needed 🔒 Unlock all questions — Premium 🤍 Like

Related Exams

Free

AWS Certified AI Practitioner (AIF-C01) - 333 Questions

333 questions · 0 attempts
⭐ Premium

AWS Certified Machine Learning Engineer Associate (MLA-C01) - 340 Questions

340 questions · 0 attempts
⭐ Premium

AWS Certified Generative AI Developer - Professional (AIP-C01) - 340 Questions

340 questions · 0 attempts
⭐ Premium

Google Cloud Generative AI Leader (GAL) - 340 Questions

340 questions · 0 attempts

Discussion

No comments yet. Be the first to start the discussion!

Sign in to join the discussion.