CCDV-F — Claude Certified Developer, Foundations
The builder’s credential. CCDV-F is for engineers who actually ship on Claude — the Messages API, agent loops, MCP servers, Claude Code — and its blueprint is unusually blunt about it: a third of the exam is applications and integration. It rewards having built something real far more than having read about it. One of four Claude certifications.
This is an independent, third-party study resource — not affiliated with, endorsed by, or an official product of Anthropic. “CCDV-F,” “Claude,” and “Anthropic” are trademarks of their owners, used here only to describe what this course helps you study. The practice questions are original study material, not real exam content. Always confirm current exam details on Anthropic’s official certification pages.
This badge is for the people who build things out of the AI, not just talk to it. Like the difference between driving a car and being the mechanic who builds one. To pass, you have to know how the engine parts fit together — and the biggest chunk of the test is about connecting your creation to everything else.
The exam guide draws a hard line around who this is for: AI/ML engineers, technical leads and senior software engineers who build and ship. It explicitly excludes non-technical Claude users, people without hands-on development experience, and roles limited to writing prompts without wider application-development responsibility. Python and/or TypeScript, REST APIs and CLI fluency are assumed.
Foxy: Thirty-three percent on “applications and integration”? That’s a third of the whole exam on one domain!
Benny the Beaver: Because that’s where real projects actually live, Foxy. Anyone can make one API call work. Making it work inside a system with auth, retries, versioning and other people’s services — that’s the job.
Pip the Hummingbird: And notice the flip side: Claude Code is only three percent, and eval and debugging under three. Don’t spend a fortnight on the small baskets.
Professor Owl: Study the blueprint before the material, class. It is telling you where the marks are.
☺ Like you’re 10: One basket on this test is nearly as big as all the little ones put together — connecting your project to other things. Put your study time where the marks are.
| Logistics | |
|---|---|
| Format | 53 items · multiple-choice and multiple-response (each item says how many to pick) |
| Time | 120 minutes |
| Pass | 720 on a scaled 100–1,000 range, set by a formal standard-setting study |
| Cost | $125 USD per attempt (retakes charged again; partner-tier discounts may apply) |
| Validity | 12 months, then a free non-proctored renewal assessment |
| Delivery | Proctored via Pearson VUE — online or test center |
| Prerequisites | None required. Recommended: 1–5 years software engineering, 6+ months hands-on with Claude or comparable LLM systems. |
Anthropic’s exam guide says there are no mandatory prerequisites and the credential is awarded on exam performance alone — that concerns knowledge. The Partner Academy certifications FAQ says certification is currently available to people at Claude Partner Network organizations, and registration needs a partner email on a recognized company domain; personal addresses will not work. That is an eligibility gate and it applies to all four Claude exams. Confirm yours before budgeting time or money.
The eight domains
☺ Like you’re 10: Eight baskets, wildly different sizes. One is a third of the test; two are barely a rounding error.
| Domain | Weight | Core of it |
|---|---|---|
| 2 · Applications & Integration | 33.1% | Requirements, systems life cycle, Claude API mechanics, software-engineering foundations, application design, configuration management |
| 5 · Model Selection & Optimization | 16.8% | LLM and technical fundamentals, model tradeoffs, cost and token management |
| 1 · Agents & Workflows | 14.7% | Agent architecture, building agents with Claude, agent patterns and frameworks |
| 6 · Prompt & Context Engineering | 11.0% | Context engineering, prompt engineering, output handling |
| 8 · Tools & MCPs | 10.6% | Tool implementation, MCP server development, agentic customization |
| 7 · Security & Safety | 8.1% | Application security, guardrails and safe deployment, hooks, identity/secrets/key management |
| 3 · Claude Code | 3.1% | Claude Code operation |
| 4 · Eval, Testing & Debugging | 2.6% | Debugging and error handling |
The Developer blueprint maps onto this course’s Applied tier almost one-for-one:
| Exam domain | Study it here |
|---|---|
| Applications & integration (33%) | Building with Claude for the API surface; AI Pipelines for wiring it into real systems; the Capstone to actually ship one end to end |
| Model selection & optimization | Models & Local LLMs for the tradeoffs; Cost, Latency & Ops for token and latency management |
| Agents & workflows | Agentic AI Concepts for the loop; Building Your Own Agents for the shapes; LLM vs RAG vs Agent vs Agentic for whether you need an agent at all |
| Prompt & context engineering | Prompting & Context, then Context Engineering and Structured Outputs & Tools |
| Tools & MCPs | MCP — Tools for Agents for building servers; Agent Protocols for where MCP stops and A2A starts |
| Security & safety | AI Security for the attack surface; Guardrails as Code for enforcing limits programmatically |
| Claude Code | Claude Code — small weight, quick win |
| Eval, testing & debugging | Evaluation & Testing and Evaluating AI Systems |
The recurring lesson of the Developer track: the model is one component in a system, and most failures are system failures. Retries, idempotency, schema validation, secret handling, versioning, graceful degradation when the model returns something unexpected — these are ordinary engineering disciplines, and the exam weights them accordingly. If a question looks like it is about Claude but is really about error handling, it probably is.
A self-contained study program (pick 2, 3, or 4 weeks) that paces you from the concepts to exam readiness, with a calendar tracker and a countdown.
Then do the Lab Track — it is the closest thing here to the hands-on experience this exam assumes, and the MCP and agent labs map directly onto two of the domains.
Build the thing the exam is really about: a small Claude-backed service that (a) calls the Messages API and handles stop_reason correctly through a full tool-use turn, (b) validates the model’s structured output against a schema and retries on failure rather than trusting it, and (c) keeps its key out of the code. Then deliberately break each part and watch what happens. That covers the two biggest domains between them.
Original practice questions across all eight CCDV-F domains, weighted roughly like the real blueprint — so most of what you see is applications, integration and model selection. Options shuffle on every attempt; drill by domain or retry only your misses.
▶ Launch the CCDV-F practice simulator
Practice questions are original study material written against the published domain blueprint — not real exam content, and not a predictor of your score.
- Official exam guide — the authoritative blueprint, per-domain objectives, policies and sample questions with rationale. Linked from the Partner Academy certification page; prefer it over any third-party summary, including this one.
- Anthropic Partner Academy — registration, the official prep course, and the certifications FAQ where the eligibility rules live: partner certifications hub.
- Pearson VUE — exam delivery and scheduling: Claude Certification Program.
- Anthropic docs — the API reference and MCP specification are the ground truth for the two biggest technical domains: docs.anthropic.com · modelcontextprotocol.io.
Certification details change, and this page is a snapshot. Confirm the fee, format, eligibility and domain weights on Anthropic’s own pages before you book. True at time of writing.
If you design systems rather than build them, look at CCAR-F or CCAR-P. If you use Claude without writing code, CCAO-F. The comparison page lays all four side by side.