Anthropic · CCDV-F
Claude Certified Developer – Foundations (CCDV-F) Practice Tests
Validates foundational knowledge and skills for developers building applications with Anthropic's Claude AI models. Covers API integration, prompt engineering, model selection, agents, workflows, and security best practices.
Exam Details
Practice Questions
302
≈ 4 practice exams
Duration
120 minutes
Passing Score
720/1000
Difficulty
FoundationalLast Updated
Sep 2026
Topics Covered
Applications & IntegrationModel Selection & OptimizationAgents & WorkflowsPrompt & Context EngineeringTools & MCPsSecurity & SafetyClaude CodeResponsible AI Development
Claude Certified Developer – Foundations (CCDV-F) Practice Exam Preparation
Use this CCDV-F practice exam to prepare for Claude Certified Developer – Foundations (CCDV-F) with realistic questions, detailed explanations, and focused study modes. The practice bank includes 302 questions for Anthropic CCDV-F, so you can review the exam steadily instead of relying on one long cram session.
As you practice, pay extra attention to recurring topics such as Applications & Integration, Model Selection & Optimization, Agents & Workflows, Prompt & Context Engineering, and Tools & MCPs. Start with short sessions to identify weak areas, then move into timed quizzes once your accuracy is consistent.
The explanations are written for candidates preparing for Anthropic certification questions and reviewing how the concepts apply in practice. Use the free preview first, then unlock the full question bank when you are ready to build a complete study routine.
Sample Questions
5 sample questions with answers and explanations. The full bank has 302 questions, enough for 4 full-length practice exams.
Preview — answers shown1. A developer is running a production agent that occasionally loops indefinitely, repeatedly calling the same tool and getting the same error. The developer wants to mitigate this issue. Which of the following is the best first mitigation? (Select one!)
AIncrease the model's temperature to encourage more varied behavior.
BAdd stop conditions that detect repeated identical failures and return actionable error messages.
CRemove all tools from the agent to prevent it from calling them.
DSwitch to a stronger model, such as Opus, to improve reasoning.
Explanation
The best first mitigation for an agent looping indefinitely is to add explicit stop conditions and improve tool-error feedback. Stop conditions can detect repeated identical failures and halt the loop, while clear error messages help the model recover. Removing all tools is too drastic and defeats the purpose of the agent. Increasing temperature or using a stronger model does not address the root cause of the loop, which is the lack of termination criteria.
2. A developer is managing a Claude application that uses multiple prompts and plugins. They need to ensure that changes are traceable and that the application can be rolled back if a prompt update causes issues. Which configuration management practice is most effective? (Select one!)
AStore prompts and plugin versions in version control and use semantic versioning
BUpdate prompts directly in the production environment and monitor performance
CKeep prompts in a shared document and manually track changes
DUse a single prompt file that is overwritten with each change
Explanation
Configuration management for prompts and plugin dependencies requires version control and semantic versioning to enable traceability and rollback. Directly updating production or manually tracking changes lacks traceability and rollback capability. Overwriting a single prompt file loses history and makes rollback impossible.
3. A developer is setting up an MCP server that will run as a separate process on the same machine as the Claude application. They need to choose a transport for local communication. Which transport is appropriate for this scenario? (Select one!)
AWebSockets, because they provide the lowest latency for local communication.
BHTTP REST, because it is the standard for all MCP servers.
CNetwork sockets, because they are the only transport supported by MCP.
Dstdio, because it is designed for locally spawned servers communicating over standard input/output.
Explanation
MCP supports stdio for locally spawned servers, where the client and server communicate over standard input/output. Network sockets are for remote servers. HTTP REST and WebSockets are not MCP transports. The scenario specifies a local process, so stdio is correct.
4. A developer is using prompt caching to reduce cost for a request that includes a long, stable system prompt. They place a timestamp at the beginning of the system prompt, before the cache checkpoint. What is the likely effect? (Select one!)
AThe cache will only be invalidated if the timestamp changes by more than one minute.
BThe cache will still work because the timestamp is small.
CThe cache will be invalidated on every request, eliminating the cost savings.
DThe cache will work, but only for the Haiku model tier.
Explanation
Placing a timestamp or user ID before the cache checkpoint invalidates the prompt cache on every request because the prefix changes, preventing reuse of the cached prefix. This eliminates the cost and latency savings. The size of the timestamp does not matter; any change before the checkpoint breaks the cache. Cache behavior is not tier-specific.
5. A developer is building a chat application using the Messages API. After sending a request, the response includes a stop_reason of 'max_tokens'. The developer's code currently parses response.content[0].text and displays it to the user as the final answer. What should the developer do to handle this correctly? (Select one!)
ADisplay the text as-is, since the API returned a 200 status code
BIgnore the stop_reason and rely on the content length to determine completeness
CCheck the stop_reason and, if it is 'max_tokens', inform the user the response was truncated and optionally continue the conversation
DRetry the same request with a higher temperature to get a complete response
Explanation
The stop_reason field must be checked before parsing content. A 'max_tokens' stop reason indicates the output was truncated because the token limit was reached, so the content should not be treated as complete. The correct action is to detect this and either inform the user or continue the conversation with additional context. Displaying the text as-is risks showing incomplete answers. Retrying with a higher temperature does not address the token limit. Relying on content length is unreliable because truncation can occur at any point.
$17.99
One-time access to this exam
302 questions (4 practice exams' worth)
Unlimited timed exam simulations
Yours forever — no subscription
Detailed explanations
Free preview stays available