
Messages API request and response anatomy, and streaming with server-sent events
Error types, retries and timeouts; images, files and thinking blocks in a request
Realtime or Message Batches API, and Claude on Amazon Bedrock, Google Cloud and Microsoft Foundry
Async calls with a limit, Claude code review in CI and safe refactoring
Tools: API Client Starter Kit for Python and TypeScript, Exam Trainer on Domain 2 skills (14.2% of the exam), flashcards, memory and match games
Functional and infrastructure requirements, success criteria and the system life cycle
How each Claude interface reads instructions: the API and SDKs, the Agent SDK, Claude Code and the Claude apps
Content boundaries, output schemas and session hygiene
Configuration files and their scopes, model pinning, prompt versions and plugin dependencies
Tools: App Design Spec with a config pack download, Exam Trainer on Domain 2 skills (18.9%) and study games
Tokens, the context window and next-token sampling; thinking, effort and fast mode
Model choice across Fable, Opus, Sonnet and Haiku by quality, latency and cost
SDK calls over the REST API and the cost formula for one Claude call
Prompt caching with 5-minute and 1-hour lifetimes, and Batch API discounts
Tools: Model and Cost Calculator, Exam Trainer on Domain 5 (16.8%) and study games
Prompt parts and where each goes; few-shot examples and output constraints
Context budget: tool output pruning, compaction and subagents
Structured output, defensive parsing, and recovery by error type and status code
Failure triage: integration layer or model output
Tools: Prompt and Output Validation Kit with a failure triage tree, Exam Trainer on Domains 6 and 4 (13.6%) and study games
Three places a tool runs: your code, Anthropic's servers or an MCP server
Tool definitions with a strict schema; a tool-use loop with dispatch, errors and an approval gate for write tools
An MCP server on the official SDK in Python and TypeScript, over stdio
Four ways to extend Claude: built-in tools, custom tools, Skills or MCP
Tools: Tool Schema Kit, FlagDesk MCP Server code pack, Exam Trainer on Domain 8 (10.6%) and study games
Workflow or agent, and the agent loop: the Agent SDK, your own harness or a framework
Subagents for context isolation and hooks for deterministic actions
Self-hosted or Anthropic-hosted agents
Claude Code: memory files, skills, subagents and headless mode
Tools: Agent Blueprint, Claude Code Config Pack, Exam Trainer on Domains 1 and 3 (17.8%) and study games
Direct and indirect prompt injection, and five input defenses
Leak, PII and identity checks; four guardrail layers from prompt to sandbox
A PreToolUse hook that blocks destructive commands, and hook callbacks in the Agent SDK
API keys across development and production workspaces
Tools: Security Kit with a 24-item checklist, hook builder and tester, Guardrail Hook Kit code pack, Exam Trainer on Domain 7 (8.1%) and study games
Eight domains, 53 items by weight, the scaled pass mark of 720 and the score report by domain
A time plan of 2.26 minutes per item and four option smells to cross out
A weighted readiness score with three zones: the course bar is above 800 with no domain under 60%
Study hours split by gap; exam-day rules, retakes and renewal
Tools: Readiness Scorecard (Excel) and the 53-question Mock Exam with a 120-minute timer, flags and a review screen
This course contains the use of artificial intelligence
A third of the CCDV-F exam is about one domain most candidates barely study: Applications and Integration.
This course follows the official Exam Guide v1.0 skill by skill, with working Python and TypeScript code, 7 exam trainers, study games and a 53-question mock exam.
You build with Claude every day, so the Claude Certified Developer – Foundations exam feels like home ground. Then the questions start: batch or realtime for a nightly job, which model tier for a support bot on a budget, where a failure starts when the JSON comes back broken, how to stop an agent from following instructions hidden in a web page. Every option sounds like something a good engineer would do. Most people prepare on agents and prompts, while API mechanics, application design and configuration carry 33.1% of the score. And a failed attempt costs you the $125 fee again plus at least 14 days of waiting before the retake.
After this course you read a question and know which of the 25 skills it tests. You call the Messages API with streaming, images, thinking and batches and handle every error type. You price a request before you ship it and cut the bill with the right model, prompt caching or a batch. You write prompts that hold, validate what comes back, build tools and an MCP server that fail safely, choose between a workflow and an agent, configure Claude Code for a repository and keep keys and injected input under control. You walk into the exam with a readiness score by domain, and you keep a developer's kit for your own Claude applications.
The course is built and taught by Mike:
Founder and head of strategy of PapaHR Academy, where 170,000+ students have trained
Author of the Claude for HR, Claude for Recruiter, Prompt Engineering for HR and HR AI Agents courses at PapaHR Academy
#1 HR instructor on Udemy, with 2,000,000+ students on Udemy
20 years of hands-on practice at Wargaming, Preply, iDeals and Starlightmedia
The course is based on the official CCDV-F Exam Guide v1.0 (July 2026) and covers all eight domains and 25 weighted skills. The heaviest domain, Applications and Integration, gets two lessons: first the mechanics of the Claude API and the engineering basics, then application design and configuration. Next you learn model selection, tokens and cost, then prompts, context, output handling and debugging. After that you build tools and a working MCP server, move on to agents with the Claude Agent SDK and Claude Code, and close the technical part with security, guardrails and hooks. Finally you calculate your readiness score and sit a full mock exam: 53 new questions in the proportions of the eight domains, a 120-minute timer, flags, a review screen and a score report by domain with a readiness zone.
Practice is where this course differs most. Every technical lesson comes with its own interactive exam trainer: 12 questions in both exam formats, one best answer and "choose two", with an instant explanation of why each wrong option is wrong and a retry of the questions you missed. Every trainer also has study games: flashcards with "know it / still learning", a memory game and a timed match game on the API parameters, SDK calls, config keys and patterns of the domain. Every practice question is written for this course; there are no real exam questions or dumps. The readiness score is the course's own estimate, not Anthropic's scoring: the official pass mark is 720 on the scaled score, and the course asks you to clear a stricter bar, above 800 with no domain under 60%, before you book. Everything interactive opens in any browser, on a phone too. All code in the course is checked against current versions of the Anthropic Python and TypeScript SDKs, the Claude Agent SDK and the official MCP SDK.
This is an independent course. It is not affiliated with, sponsored by or endorsed by Anthropic.
What's included:
Lifetime access to all materials
Active instructor support in Q&A
Udemy Certificate of Completion
A practical assignment in every lesson, built on your own Claude project in lessons 1–7
7 interactive exam trainers with 84 practice questions, 12 per lesson, and explanations
Study games in every trainer: flashcards, memory and timed match
A 53-question mock exam with a 120-minute timer, flags, a review screen and a score report by domain
A readiness scorecard in Excel: weighted score by domain, readiness zone and study hours by gap
7 practice tools that run offline in the browser: API Client Starter Kit, App Design Spec, Model and Cost Calculator, Prompt and Output Validation Kit, Tool Schema Kit, Agent Blueprint, Security Kit
3 code packs in Python and TypeScript: FlagDesk MCP Server, Claude Code Config Pack, Guardrail Hook Kit
A section with additional courses, tools and resources
The Claude API, the Agent SDK and Claude Code change every few months, and the credential is valid for 12 months for exactly that reason. The developers who can show it now are the ones their teams will trust with the next Claude build. Every week you wait is another week of guessing on questions you could already answer.
Enroll now and start your first lesson today.