Claude Certified Developer — Foundations (CCDV-F): complete study guide

The CCDV-F certifies that you can build software on Claude — call the API correctly, choose a model deliberately, wire up tools and MCP servers, decide between a workflow and an agent, and ship something that behaves sensibly when things go wrong. It sits between the Associate exam, which is about using Claude well, and the Architect exams, which are about designing multi-agent systems.

What this exam is for

The CCDV-F certifies that you can build software on Claude: call the API correctly, choose a model deliberately, wire up tools and MCP servers, decide between a workflow and an agent, and ship something that behaves sensibly when things go wrong. It sits between the Associate exam, which is about using Claude well, and the Architect exams, which are about designing multi-agent systems.

The audience is technical builders — AI and ML engineers, technical leads and senior software engineers. The expected background is one to five years of software engineering, at least six months hands-on with Claude or a comparable LLM, working Python or TypeScript, and comfort with REST APIs and a command line.

Format

Note the multiple response format: some questions ask you to select two or three correct answers, and partial credit is not typically awarded. Read the instruction line on every question rather than assuming one answer.

Domain weights

The distribution is unusually lopsided, and it should drive your preparation directly. Domain 1 alone is a third of the exam — more than domains 4 through 8 combined. Domains 1 to 3 together are nearly two thirds. Claude Code, which dominates the Architect exam, is worth about two questions here; do not spend a week on it.

How the questions are built

Expect code-shaped scenarios: a request that returns something unexpected, an integration that fails intermittently, a cost that is too high, a pipeline that needs to be reliable. The distractors are usually plausible engineering, failing on one specific point — a field misread, a retry policy applied to a permanent error, an abstraction reached for too early.

Three habits pay for themselves across the whole paper: read what the API actually returns rather than what you assume it returns; prefer the simplest construction that meets the requirement; and check whether the stem's constraint is latency, cost, correctness or reliability, because that is usually what separates the two most attractive options.

Every page in this guide

• The mental model — One stance underlies most CCDV-F questions: the API gives you a contract and the model gives you a guess. Branch on the contract, verify the content, and start simple.
• Domain 1 — Applications and Integration — A third of the whole exam: the Messages API request and response, stop reasons, streaming, multimodal input, retry policy and configuration management.
• Domain 2 — Model Selection and Optimisation — Tokens and context windows, choosing a model tier deliberately, thinking budgets, and the cost levers — prompt caching, batching, routing and rate limits.
• Domain 3 — Agents and Workflows — Workflow versus agent, the five workflow patterns, autonomous agents, and building with the Claude Agent SDK.
• Domain 4 — Prompt and Context Engineering — Prompt technique in order of value, context engineering and context editing, structured output, and validating responses with error-specific retries.
• Domain 5 — Tools and MCPs — The tool-use exchange, designing tools a model can actually use, MCP architecture and primitives, and building an MCP server.
• Domain 6 — Security and Safety — Direct and indirect prompt injection, layered defences, guardrails, secrets management and PII handling.
• Domain 7 — Claude Code — Where configuration lives, what enforces versus what only guides, permission modes and running Claude Code in CI.
• Domain 8 — Eval, Testing and Debugging — Isolating integration failures from model-output failures, building evals, the three grading methods and regression discipline.
• The 13 anti-patterns — Thirteen distractors that recur across the CCDV-F question bank, each with why it fails and what the exam wants instead.
• Cheat sheet — Stop reasons, the retry matrix, cache and batch limits, the five workflow patterns, schema support and the MCP quick reference.
• Questions & answers — Domain 1 — CCDV-F practice questions on applications and integration, each with the correct answer and a full explanation of why the distractors fail.
• Questions & answers — Domain 2 — CCDV-F practice questions on model selection and optimisation, each with the correct answer and a full explanation of why the distractors fail.
• Questions & answers — Domain 3 — CCDV-F practice questions on agents and workflows, each with the correct answer and a full explanation of why the distractors fail.
• Questions & answers — Domain 4 — CCDV-F practice questions on prompt and context engineering, each with the correct answer and a full explanation of why the distractors fail.
• Questions & answers — Domain 5 — CCDV-F practice questions on tools and mcps, each with the correct answer and a full explanation of why the distractors fail.
• Questions & answers — Domains 6–8 — CCDV-F practice questions on prompt injection defences, secrets and PII, Claude Code configuration and debugging, with answers and explanations.
• Facts to memorise — Thirty recall facts covering stop reasons, the retry matrix, cache and rate-limit mechanics, the workflow patterns and MCP rules.
• Five-week study plan — A week-by-week CCDV-F plan that spends two weeks on Applications and Integration because it is a third of the exam, and hours on the light domains.
• Official resources — The Messages API reference, rate limits, prompt caching, building effective agents, MCP docs and jailbreak mitigation, mapped to the eight domains.

This is an independent study guide, not an official Anthropic publication. Exam blueprints and API details change; confirm the current domain weights and format against the official exam guide before you sit.

Next: The mental model →