Claude Certified Developer Foundations (CCDV-F) : Exam Prep

• Allocate 120 minutes across 53 multiple-choice and multiple-response items and apply elimination strategy to plausible distractors
• Select the correct API mechanic – streaming, batch, caching, or error recovery – for items drawn from Applications and Integration, the exam's heaviest domain at 33.1%
• Diagnose the anti-patterns most frequently tested across the eight published domains and select the correct remediation
• Distinguish deterministic enforcement through hooks, schemas, and guardrails from prompt-based guidance when an item asks for the most reliable solution
• Select among built-in tools, custom tools, Skills, and MCP servers for a stated requirement under exam conditions
• Answer full-length timed item sets at exam pace, weighted to the published domain percentages
• Evaluate personal readiness by domain using a scored gap analysis weighted to the published domain percentages
• Construct a targeted study plan addressing the domains where measured performance falls short of the passing standard
• Schedule the exam against the published registration, identification, retake, and proctoring policies

CCDV-F certification candidates who have completed the Developer track or have equivalent hands-on experience across the eight exam domains. AI and machine learning engineers, technical leads, and senior software engineers who have built with the Claude API, custom tools, MCP servers, and the Claude Agent SDK and now need timed, exam-format practice rather than further instruction.
 

The Claude Developer Intensive course, or demonstrated equivalent experience. Candidates should have hands-on experience with the Claude API, tool use and function calling, MCP server development, the Claude Agent SDK, and Claude Code. Anthropic's stated candidate profile assumes one to five years of software engineering experience, at least six months of hands-on work with Claude or a comparable LLM platform, and proficiency in Python and/or TypeScript. The exam itself has no mandatory prerequisites. Roughly a fifth of the published blueprint tests general software-engineering and requirements competence rather than Claude-specific skill; this course reviews that material as an exam surface and assumes candidates already hold the underlying skill.
 

Exam Structure, Format, and Strategy

• Format: 53 items in 120 minutes; multiple-choice and multiple-response, with each item stating how many responses to select
• Scaled scoring from 100 to 1,000 with a cut score of 720; percent-correct by domain appears on the score report but does not determine the result
• Criterion-referenced standard: the comparison is against the minimally qualified candidate, not the cohort
• Reading an item stem for the decision it actually asks you to make
• Pacing at roughly two minutes per item, and where to spend the surplus
• Drill: pacing calibration against a mixed item set

Domain 1 Review: Agents and Workflows (14.7%)

• Workflow versus agent: the decision criteria the exam tests, and manager and supervisor hierarchies
• Subagents and their role in improving task execution
• Agent construction: the Claude Agent SDK, custom agent loops and harnesses, and self-hosted versus Anthropic-hosted deployment
• Hooks for deterministic actions inside an agent
• Agent patterns: tool-use loops, memory, and context-window management
• Agentic abstraction frameworks named in the blueprint: Strands, LangGraph, PydanticAI
• Drill: timed item set proportional to the domain's 14.7% share

Domain 2 Review: Applications and Integration (33.1%)

• Claude API mechanics: messages, tools, streaming, vision, thinking, caching, and Messages API data access patterns
• Realtime versus batch: the Message Batches API for latency-tolerant, high-volume work, and the cost tradeoff behind the choice
• Invoking Claude through third-party vendors, and what changes when you do
• Software engineering foundations under exam conditions: REST, JSON, asynchronous programming, version control, SDLC integration, code review, and refactoring at small and large scale — assumed skill, reviewed here only as an exam surface
• Claude application design: how Claude interprets instructions across Claude Code, Desktop, Claude.ai, the API, and the SDKs; content boundaries; schema design; session hygiene; plugin management
• Configuration management: CLAUDE.md, settings.json, model version pinning, prompt versioning, and plugin dependencies
• Requirements and systems life cycle: deriving functional and infrastructure requirements from business requirements
• Drill: extended timed item set proportional to the domain's 33.1% share, the largest single block of the day

Domain 3 Review: Claude Code (3.1%)

• Core components: Rules, Skills, Commands, Agents, and Agent Memory
• Features: session management, built-in and custom slash commands, headless mode, streaming mode, and auto-mode
• The CLAUDE.md hierarchy, repository initialization, and settings.json configuration
• Drill: short timed item set; at 3.1% of the exam this domain is worth minutes, not hours, and the day is budgeted accordingly

Domain 4 Review: Eval, Testing, and Debugging (2.6%)

• Error type identification and recovery strategy selection
• Trace analysis to identify failure modes
• Isolating problem origin between the integration layer and model output
• Drill: short timed item set; the published blueprint scores this domain through debugging and error handling alone

Domain 5 Review: Model Selection and Optimization (16.8%)

• LLM fundamentals: tokens, context windows, sampling, non-determinism, and next-token generation
• Model options: fast mode, extended thinking, adaptive thinking, and effort levels
• Zero-shot, single-shot, and multi-shot prompting as the blueprint frames them
• Opus, Sonnet, and Haiku: quality, latency, and cost tradeoffs, and breaking behavior changes across model releases
• Cost and token management: usage tracking, cost modeling, prompt caching, and cache check-pointing
• Technical fundamentals: SDKs that wrap REST APIs, and websockets
• Drill: timed item set proportional to the domain's 16.8% share

Domain 6 Review: Prompt and Context Engineering (11.0%)

• Context engineering: context-window management, context drift and bloat, tool output pruning, and compaction
• Context isolation through subagents and multi-step agentic workflows
• Prompt engineering: instruction clarity, few-shot examples, system versus user placement, output constraints, iterative refinement, and input sanitization
• Output handling: structured output patterns, response validation, defensive parsing, and skepticism toward confident output
• Drill: timed item set proportional to the domain's 11.0% share

Domain 7 Review: Security and Safety (8.1%)

• Prompt injection and jailbreak defense: isolating untrusted input from trusted instructions, and why a polite system-prompt request is not a control
• Data leakage prevention, PII handling, and authentication, authorization, confidentiality, and integrity
• Guardrails and safe deployment: content policy, guardrail layering, and secure-by-design least privilege
• Claude hooks as the deterministic control that prevents destructive actions
• Identity, secrets, and key management across development and production environments
• Drill: timed item set proportional to the domain's 8.1% share, including the published sample item on hidden instructions in retrieved web content

Domain 8 Review: Tools and MCPs (10.6%)

• Tool implementation: function calling, tool descriptions, error handling, and tool set construction
• Tool usage patterns: agentic harness dispatch, client-side versus server-side tools, and approval patterns
• MCP server development: authoring, deployment, resources, tools, prompts, and communication patterns (stdio, sockets, client versus server)
• Agentic customization: the tradeoffs among built-in tools, custom tools, Skills, and MCPs for a given use case
• Drill: timed item set proportional to the domain's 10.6% share

Timed Drills and Item Debrief

• Full-length simulation: 53 items in 120 minutes under exam conditions
• Item mix drawn in proportion to the published domain weights, with a third of the items in Applications and Integration
• Multiple-response items: reading the stated number of correct responses and abandoning partial-credit assumptions
• Item-by-item debrief: why the keyed answer is correct and why each distractor is attractive
• Anti-pattern catalog: the errors the exam consistently rewards candidates for rejecting
• (Anthropic publishes no scenario bank for this exam; drills use LearnQuest production scenarios written to the published blueprint)

Personal Gap Analysis and Study Planning

• Scored gap analysis across the eight domains, weighted to the published domain percentages
• Mapping measured gaps back to CL-250 material and to Anthropic's published preparation guidance
• Building and operating one application that exercises the API, tools, prompt and context engineering, security, and evaluation, as the exam guide recommends
• Building a targeted personal study plan with a concrete exam date
• Registration, identification, proctoring, retake, and renewal policy walkthrough