The correct answer is highlighted above.
More Information
| Anthropic | Anthropic |
|---|---|
| Exam name | Claude Certified Developer – Foundations |
| Exam Code | CCDV-F |
| Number of Question | 95 |
| Update at | Nov 6, 2026 |
Description
Claude Certified Developer – Foundations (CCDV-F)
Everything you need to master the CCDV-F exam — structure, all eight blueprint domains, fees, and tactical study plans, paired with code-grounded Claudemock™ practice simulations.
01. What is the CCDV-F Certification?
The Claude Certified Developer – Foundations (CCDV-F) is Anthropic's hands-on engineering credential for building on Claude. Built for software engineers, backend and full-stack developers, and AI engineers, it assesses your ability to integrate the Claude API into real applications, build agents and tool-calling loops, choose and optimize models, and ship with sound security and evaluation practice.
Administered via Pearson VUE (online proctoring through OnVUE or physical testing centers), passing candidates earn a verified digital Credly badge valid for 12 months, renewable through a free, non-proctored annual assessment.
CCAO-F
Foundations for day-to-day business productivity, prompting, & data governance.
CCDV-F
API integration, Agent SDK implementation, & programmatic tool-calling.
CCAR-F
Foundational agentic architecture, MCP tool design, Claude Code, & reliability.
CCAR-P
Advanced multi-tenant systems, high-scale agent orchestration, & governance.
02. Exam Format & Specifications
| Questions | 53 items (Multiple-choice & multiple-response), sampled directly from the blueprint |
|---|---|
| Duration | 120 minutes (~2.25 minutes per item) |
| Passing Benchmark | Scaled score of 720 (range: 100–1,000) |
| Testing Delivery | Pearson VUE OnVUE (remote proctored) or Authorized Test Centers |
| Registration Fee | $125 USD per attempt (No complimentary retakes) |
| Prerequisites | Corporate association with a Claude Partner Network tier (Free "Registered" tier eligible). Working Python or TypeScript and Claude API experience recommended |
| Cooldown / Retake Policy | Attempt 2: 14-day wait | Attempt 3: 30 days | Attempt 4: 90 days (Max 4/year) |
| Validity & Renewal | 12 months; free, non-proctored annual renewal module |
03. Exam Domains & Blueprint Weights
Half the exam (49.9%) sits in just two domains: Applications & Integration and Model Selection & Optimization. Prompt wording alone will not carry you — you must know API mechanics, cost and latency trade-offs, and how production code handles Claude's responses.
| Domain | Weight | Core Evaluated Skills |
|---|---|---|
| 1. Agents & Workflows | 14.7% | Workflow vs. agent choice, Claude Agent SDK, custom agent loops, subagents, hosted vs. self-hosted deployment. |
| 2. Applications & Integration | 33.1% | Messages API, streaming, vision, extended thinking, prompt caching, Message Batches API, stop_reason handling. |
| 3. Claude Code | 3.1% | CLAUDE.md configuration, Skills, slash commands, hooks, headless automation. |
| 4. Evaluation, Testing & Debugging | 2.6% | Identifying API error types, eval design, graders, regression testing, diagnosing model behavior. |
| 5. Model Selection & Optimization | 16.8% | Opus vs. Sonnet vs. Haiku trade-offs, token usage tracking, cost modeling, latency optimization. |
| 6. Prompt & Context Engineering | 11.0% | System prompts, XML structure, few-shot examples, context rot, RAG, just-in-time context. |
| 7. Security, Safety & Guardrails | 8.1% | Prompt injection defense, jailbreak mitigation, data privacy, output guardrails. |
| 8. Tools, Function Calling & MCP | 10.6% | Tool schemas, tool_choice, parallel tool calls, building MCP servers, client integration. |
04. Ideal Candidate Profile
Built For
- Software engineers integrating the Claude API into products.
- Backend and full-stack developers shipping AI features.
- AI engineers building agents with the Claude Agent SDK.
- Developers writing tools and MCP servers for Claude.
Out of Scope (Choose CCAO-F or CCAR)
- Business users relying on Claude for daily productivity (CCAO-F).
- Architects owning multi-agent system design decisions (CCAR-F).
- Leads governing high-scale, multi-tenant platforms (CCAR-P).
05. Why Candidates Fail & How Claudemock Solves It
Studying the Wrong Domains
Candidates over-prepare prompting (11%) and under-prepare API integration (33.1%).
API Features You Rarely Use
Batches, caching, tool_choice, and stop_reason handling appear even if your job never touches them.
Closed-Book Format
No docs, no IDE, no AI assistant — API behavior must come from recall.
06. Recommended 4-Week Study Sprint
Core API & Foundations
Read the official CCDV-F exam guide. Complete Anthropic Academy's Building with the Claude API, Introduction to Model Context Protocol, and Claude Code in Action. Set up an API key, SDK, and scratch project.
Domain-by-Domain Drilldowns
Complete all practice modules on Claudemock.com, prioritizing Applications & Integration and Model Selection & Optimization. Read all feedback rationales, especially for correct answers.
Hands-on Build
Ship one small app end-to-end: streaming responses, a tool-calling loop, prompt caching, a Message Batches job, a simple MCP server, prompt-injection guardrails, and a basic eval set.
Simulated Full Mocks & Pacing
Complete two timed, closed-book Claudemock exams in single sittings. Verify your domain radar chart stays consistently above 85% before scheduling on Pearson VUE.
07. Sample Examination Questions
Original practice questions matching official CCDV-F depth and difficulty.
A team must summarize 40,000 archived documents. Results are needed by the next morning, nothing is user-facing, and cost is the main constraint. Which approach fits best?
Correct Answer: B.
The Message Batches API is built for large, non-urgent workloads: it processes requests asynchronously at a significant discount. Streaming and parallelism optimize latency, which this task does not need, and one giant request exceeds practical context limits.
Your application receives a response with stop_reason: "tool_use". Which two things must the next request include for the conversation to continue correctly? (Select 2)
Correct Answers: A and C.
The API is stateless: your code runs the tool, keeps the assistant's tool_use turn in history, and returns the output as a tool_result tied to the same tool_use_id. A tool_use stop is not a token-limit problem, and disabling tools would break legitimate follow-up calls.
A support app sends the same 20,000-token product manual at the start of every request, followed by a short customer question. Input costs and time-to-first-token are too high. What is the most effective first change?
Correct Answer: B.
A large, unchanging prefix reused across requests is the textbook case for prompt caching, which cuts both input cost and latency on cache hits. max_tokens limits output, not input, and a larger model raises cost.
08. Frequently Asked Questions
Ready to Pass CCDV-F on Your First Attempt?
Sharpen your timing, lock in API mechanics, and track domain-level readiness across all eight domains with Claudemock™ practice exams.
Reviews
CCDV-F Real Exam Questions
Your Claude application's API keys are stored in a secrets manager. The team is debating whether the same key should be used in development, staging, and production environments. How would you handle the keys across environments?
Your Claude application returns confident-sounding answers, but occasionally those answers contain factual errors that downstream systems treat as ground truth. The team is concerned about the application's confidence-versus-accuracy gap. How would you address the gap?
A teammate has asked you to explain the difference between context engineering and prompt engineering. They have heard the terms used interchangeably and are unsure how each applies to a Claude application that processes long-running multi-step tasks. How would you describe the distinction?
You are choosing between using STDIO-based communication and HTTP-based communication for an MCP server. The server will be invoked by a Claude Code session running locally. Which communication pattern would you use?
A teammate has asked why your Claude application sometimes produces a response that includes the prompt text repeated back, and other times produces a response with the prompt text rephrased. They suspect a bug in the application's request construction. How would you respond?