Claude Certified Developer – Foundations (CCDV-F)

Learning plan

About 40 hours of focused study, split into 5 phases. Pick your weekly pace and the timeline adjusts.

Estimated finish: 4 weeks

What the exam covers

Source: official exam guide. Weights can change, so check before you start.

Before you start

• Eligibility: the exam is limited to people at Claude Partner Network organizations. Register via the Anthropic Partner Academy and sit it with Pearson VUE
• Proficiency in Python or TypeScript, plus comfort with REST, JSON, async code and git
• A Claude API key with a small credit balance for the labs

1. Weeks 1–2 · 13 h

   Claude API mechanics and application design (a third of the exam)

   • Call the Messages API with system prompts, multi-turn history, streaming, image input and extended thinking, and read usage and stop_reason from the response
   • Choose real-time vs Message Batches for a workload, and know Claude is also reachable through Amazon Bedrock and Google Cloud Vertex AI
   • Manage configuration with CLAUDE.md, settings.json, pinned model versions and versioned prompts
   • Apply software-engineering basics the blueprint tests directly, including REST error handling, async concurrency, code review and safe refactoring

   Hands-on: Build a small CLI that summarises PDFs and images with streaming output. Add retry with backoff on 429/529 errors, pin the model version in config, then run the same 20 files through the Message Batches API and compare cost.

   Use: Building with the Claude API (Anthropic Academy) · Claude API overview · Batch processing (Claude docs) · anthropics/courses on GitHub

2. Week 2 · 7 h

   Model selection, tokens and cost

   • Match Opus, Sonnet and Haiku to tasks and explain the quality/latency/cost trade-off and when extended or adaptive thinking is worth it
   • Explain tokens, context windows, sampling and non-determinism, and how effort settings change output
   • Place cache breakpoints for prompt caching and estimate the monthly cost of a workload from token counts

   Use: Claude models overview · Prompt caching (Claude docs)

3. Week 3 · 10 h

   Agents, tools and MCP servers

   • Decide workflow vs agent, and build an agent with the Claude Agent SDK including subagents and hooks for deterministic actions
   • Know where third-party frameworks (Strands, LangGraph, PydanticAI) fit, and self-hosted vs Anthropic-hosted agent deployment
   • Write tool schemas and descriptions, handle tool errors, and distinguish client-side from server-side tools and approval patterns
   • Build an MCP server exposing tools, resources and prompts over stdio, and choose built-in tools vs custom tools vs Skills vs MCP for a use case

   Hands-on: Wrap a public REST API (e.g. a weather or GitHub API) as an MCP server with two tools and one resource. Connect it to an Agent SDK agent and add a hook that blocks any tool call containing a delete verb.

   Use: Claude Agent SDK overview · Introduction to Model Context Protocol (Anthropic Academy) · Introduction to agent skills (Anthropic Academy) · Tool use with Claude (docs)

4. Week 4 · 6 h

   Prompts, context, security, Claude Code and debugging

   • Use system vs user placement, few-shot examples and output constraints, and prevent context drift with tool-output pruning and compaction
   • Treat retrieved and user content as untrusted, keep it apart from instructions, and use hooks or least privilege so prompt injection can't trigger sensitive actions
   • Keep API keys in a secrets manager or environment variables and scope them per environment
   • Know Claude Code components (rules, skills, commands, agents, memory), headless mode and settings.json, and isolate whether a bug sits in the integration layer or the model output

   Use: Prompt engineering overview (Claude docs) · Claude Code Hooks · Claude Code in Action (Anthropic Academy)

5. Week 4 · 4 h

   Sample questions and weighted review

   • Answer the guide's three sample items (batch API, prompt injection, MCP reuse) and explain each distractor
   • Re-test yourself on Applications and Integration and Model Selection, which together are about half the exam
   • Self-rate every skill in the blueprint and close any gap rated below 4/5

   Use: CCDV-F official exam guide (PDF) · CCDV-F prep path (Anthropic Partner Academy)

You're ready to book when…

• You can build a streaming, tool-using Claude app with retries and caching without copying from docs
• You can explain when to use built-in tools, custom tools, Skills or MCP for a given requirement
• You can name a concrete defence for prompt injection, leaked keys and destructive tool calls
• You get all official sample questions right and rate every blueprint skill 4/5 or higher

Exam tips

• Weights are uneven. Applications and Integration is 33.1% while Claude Code is 3.1% and Eval/Debugging 2.6%, so budget time to match.
• The exam has 53 items in 120 minutes with a pass mark of scaled 720/1000. Multiple-response items tell you how many answers to pick.
• Updated exam content takes effect December 7, 2026 (new guides November 17, 2026), so re-check the blueprint if you book after that.