Claude Certified Developer — Foundations (CCDV-F)
What this chapter is. A complete, self-contained study course for the Claude Certified Developer – Foundations exam. Written to teach: intuition → mechanism → runnable example → the decision a developer must make. Read in order for a first pass, then drill
PRACTICE_QUESTIONS.md.
⚠️ Accuracy note. Numbers below are cross-checked against public study guides as of August 2026; confirm against the official Exam Guide PDF on the Anthropic Partner Academy (Skilljar) before booking. Product behavior reflects current Anthropic docs (
platform.claude.com,code.claude.com).
Relationship to the Architect cert (
../71_claude_certified_architect/). The two exams overlap heavily (API, MCP, Claude Code, agents). The Architect exam tests design trade-offs; the Developer exam tests implementation — writing the code, choosing models, handling errors, streaming, caching, security. Where a concept is fully developed in the Architect chapter, this one links to it and focuses on the developer’s hands-on angle. If you’re taking both, study them together.
1. What the certification tests
A certified developer can build applications with Claude: integrate the API, choose and optimize models, wire tools and MCP servers, engineer prompts and context, and handle security and debugging — in code.
The tell that this is a developer exam: Applications & Integration is 33.1% — by far the heaviest single domain on any Claude certification. Expect concrete questions about messages, tools, streaming, vision, caching, batching, error handling, and async patterns.
2. Exam logistics (cross-checked; confirm before booking)
| Item | Value |
|---|---|
| Exam code | CCDV-F (a.k.a. CCD-F) |
| Questions | 53 scored |
| Time | 120 minutes (~2.3 min/question) |
| Passing score | 720 / 1000 (scaled) |
| Question types | Multiple-choice and multiple-response |
| Cost | $125 USD |
| Delivery | Pearson VUE — online-proctored or test center |
| Validity | 12 months |
| Level | Foundations (entry tier) |
3. Domain weightings — spend time by weight
| # | Domain | Weight | This chapter’s file |
|---|---|---|---|
| 1 | Applications & Integration | 33.1% | APPLICATIONS_INTEGRATION_DEEP_DIVE.md |
| 2 | Model Selection & Optimization | 16.8% | MODEL_SELECTION_OPTIMIZATION.md |
| 3 | Agents & Workflows | 14.7% | AGENTS_AND_WORKFLOWS.md |
| 4 | Prompt & Context Engineering | 11.0% | PROMPT_AND_CONTEXT_ENGINEERING.md |
| 5 | Tools & MCPs | 10.6% | TOOLS_AND_MCP.md |
| 6 | Security & Safety | 8.1% | SECURITY_AND_SAFETY.md |
| 7 | Claude Code | 3.1% | CLAUDE_CODE_AND_DEBUGGING.md |
| 8 | Evaluation, Testing & Debugging | 2.6% | CLAUDE_CODE_AND_DEBUGGING.md |
Domains 1 + 2 are ~50% of the exam. If time is short, master API integration and model/cost optimization first.
4. Six things a Claude developer must be fluent in
- The Messages API request/response shape — roles,
system, content blocks,stop_reason, token accounting. Stateless: you resend history. - Tool use loop in code — send tool defs → get
tool_use→ run it → returntool_result→ continue. - Model trade-offs — Opus vs Sonnet vs Haiku on capability/latency/cost; when extended thinking pays off.
- Cost/latency levers — prompt caching, batching, streaming, token/context management, model routing.
- Reliability in code — structured outputs, error handling, retries with backoff, idempotency, async.
- Security — prompt injection defense, untrusted-input handling, PII, key management, least privilege, hooks.
5. How to study this chapter
Pass 1 — read in order: APPLICATIONS_INTEGRATION_DEEP_DIVE.md → MODEL_SELECTION_OPTIMIZATION.md → AGENTS_AND_WORKFLOWS.md → PROMPT_AND_CONTEXT_ENGINEERING.md → TOOLS_AND_MCP.md → SECURITY_AND_SAFETY.md → CLAUDE_CODE_AND_DEBUGGING.md.
Pass 2 — build: run the Quickstart, make a tool-use call, add streaming, add prompt caching, and read back the usage field to see the token/cost effects. Nothing cements the integration domain like one real script.
Pass 3 — drill: PRACTICE_QUESTIONS.md, cold, explaining every distractor.
6. Suggested one-week plan
| Day | Focus |
|---|---|
| 1–2 | Applications & Integration (the 33%) + build a real script with tools, streaming, caching |
| 3 | Model selection & optimization; measure token usage and cost |
| 4 | Agents & workflows; build a small tool loop / subagent |
| 5 | Prompt & context engineering + Tools & MCP |
| 6 | Security & safety + Claude Code + eval/debugging |
| 7 | Practice questions; review misses; re-read weak files |
7. Further reading (authoritative first)
- Anthropic Partner Academy (official exam/enrollment) —
https://anthropic-partners.skilljar.com - Claude API docs —
https://platform.claude.com/docs - Client SDKs (Python/TS) —
https://platform.claude.com/docs/en/api/client-sdks - Agent SDK / Claude Code —
https://code.claude.com/docs - Prompt engineering —
https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/overview - Model Context Protocol —
https://modelcontextprotocol.io