CLAUDE · CCDV-F

Claude Certified Developer, Foundations practice questions

Engineers building on the Claude API, Agent SDK and MCP

Pass mark720 / 1000
Questions in the exam53
Time allowed120 min
Exam fee$125
Practice questions here53

What the CCDV-F exam covers

July 2026 guide
Agents and workflows14.7%
Applications and integration33.1%
Claude Code3.1%
Evaluation, testing and debugging2.6%
Model selection and optimisation16.8%
Prompt and context engineering11%
Security and safety8.1%
Tools and MCP10.6%

Sample CCDV-F questions

Five of the 53 practice questions, each with its answer and explanation.

Applications and integration

A Messages API response comes back with stop_reason 'tool_use'. What should your application do next?

  1. Show the response to the user as the final answer
  2. Run the requested tool and send a new user message containing a tool_result block with the matching tool_use_id
  3. Resend the original request unchanged
  4. Raise max_tokens and try again
Show the answer

B. Run the requested tool and send a new user message containing a tool_result block with the matching tool_use_id

'tool_use' means Claude is waiting for a tool result. Your code executes the tool and returns the result linked by tool_use_id so Claude can continue.

Applications and integration

A response stops mid-sentence and stop_reason is 'max_tokens'. What does this mean?

  1. The API returned an error
  2. The output reached the max_tokens limit you set, so raise the limit or continue the response
  3. The model refused the request
  4. The prompt was too short
Show the answer

B. The output reached the max_tokens limit you set, so raise the limit or continue the response

'max_tokens' is a normal stop reason. The reply was cut off at your limit, not rejected.

Applications and integration

Users say your chat app feels slow, although total generation time is acceptable. Which change addresses this most directly?

  1. Stream the response so text appears as it is generated
  2. Send every request twice
  3. Remove the system prompt
  4. Lower the temperature
Show the answer

A. Stream the response so text appears as it is generated

Streaming does not shorten total time, but users see the first words almost at once, which is what makes an interface feel responsive.

Applications and integration

Where do standing instructions for the model belong in a Messages API request?

  1. In the top-level system parameter
  2. In a message with the role 'system' inside the messages array
  3. In the model name
  4. In the stop_sequences list
Show the answer

A. In the top-level system parameter

The Messages API takes the system prompt as a top-level parameter. The messages array holds user and assistant turns.

Model selection and optimisation

Every request repeats the same long system prompt and reference documents. How do you cut cost and latency?

  1. Put the changing user question first and the documents last
  2. Use prompt caching, with the stable content at the start of the prompt and marked as cacheable
  3. Send the documents in a separate request that is never referenced
  4. Increase max_tokens
Show the answer

B. Use prompt caching, with the stable content at the start of the prompt and marked as cacheable

Prompt caching reuses a stable prefix across requests. It only works if the unchanged content comes first, before the parts that vary.