FlashGenius Logo FlashGenius
2026 Interactive Exam Guide · CCDV-F

Claude Certified Developer
– Foundations

Everything you need to plan for the CCDV-F exam — the official blueprint, cost, format, a 6-week study plan, common mistakes, and a personalized readiness quiz.

53
Exam Items
120 min
Time Limit
$125
Exam Fee
720/1000
Passing Score

By the FlashGenius team · Last updated July 13, 2026 · Sourced from Anthropic's official CCDV-F Exam Guide v1.0

What Is the CCDV-F Credential?

Anthropic expanded its Claude Certification Program in July 2026 to include four role-based credentials. The Claude Certified Developer – Foundations (exam code CCDV-F) validates that you can build, integrate, and ship production-grade applications, agents, and workflows on the Claude platform at a foundational level.

🔌

Integrate

API, SDKs, streaming, error handling

🤖

Build

Agents, workflows, tools, MCP servers

🛡️

Secure

Guardrails, hooks, injection defense

⚙️

Optimize

Model selection, tokens, cost, caching

What earning it signals

According to the official exam guide, this credential tells employers, clients, and teams that the holder can independently own or significantly contribute to building, integrating, and shipping Claude-powered systems — not just write prompts in isolation.

  • Building agents and workflows with the Claude Agent SDK and custom agent loops
  • Integrating Claude into application code via the API, client SDKs, and third-party vendors
  • Operating Claude Code, authoring Skills, and configuring CLAUDE.md / settings.json
  • Prompt and context engineering to control behavior and prevent context drift
  • Designing evals, debugging via trace analysis, validating structured output
  • Selecting model tiers and managing tokens, caching, and batch processing
  • Applying secure-by-design principles and hooks against prompt injection
  • Building custom tools, function schemas, and MCP servers
🗺️
Part of a four-credential program. Anthropic's Claude Certification Program also includes Claude Certified Associate – Foundations, Claude Certified Architect – Foundations, and Claude Certified Architect – Professional. See the "Who It's For" tab to confirm Developer is the right path for you.

Getting Started Checklist

A concrete, clickable path from "I just found out this exam exists" to exam day. Check off steps as you complete them.

0 of 8 steps completed

1
Pre-Study

Confirm you're the right candidate

CCDV-F targets AI/ML engineers, technical leads, and senior software engineers with 1–5 years of software engineering experience and at least 6 months hands-on with Claude or a comparable LLM system. Check the "Who It's For" tab if you're unsure.

2
Pre-Study

Download and read the full official exam guide

It covers the blueprint, scoring, policies, and rules of conduct. It's the authoritative reference — read it before you build a study plan around any third-party source.

3
Pre-Study

Self-assess against each of the 8 domains

Go through the blueprint in the "CCDV-F Exam" tab and rate your comfort with each domain honestly. Applications & Integration (33.1%) and Model Selection & Optimization (16.8%) alone make up half the exam.

4
Hands-On

Build and ship one real Claude application

The exam guide explicitly recommends this. Build something that exercises the API, integrates at least one tool, applies basic prompt/context engineering, and includes simple security and evaluation practices.

5
Hands-On

Get hands-on with the Agent SDK, Claude Code, and MCP

Author a Claude Skill, configure a CLAUDE.md and settings.json, and build or connect to at least one MCP server. These are named explicitly in the blueprint and are easy to under-practice if you only read docs.

6
Exam Prep

Work through sample questions and practice items

The official guide includes illustrative sample items per domain. Study the rationale for each answer — the exam rewards understanding tradeoffs, not memorizing facts.

7
Register

Register through the Anthropic Partner Academy + Pearson VUE

Review the Certification Terms and the Certification Exam Policy before you pay. You'll create a Pearson VUE account to schedule either online proctoring or a test-center session.

8
Exam Day

Sit the exam and review your domain score report

You'll get a pass/fail result with a scaled score (100–1000, cut score 720) plus percent-correct by domain — useful for a retake plan if needed.

Who CCDV-F Is For — and How It Compares

Anthropic's Claude Certification Program has four credentials. Pick the tab that matches how you'll use Claude day-to-day to confirm Developer is the right one before you register.

Claude Certified Associate – Foundations

For client-facing practitioners: prompting, output evaluation, workflow design, responsible use, and supervised delivery scenarios. No production build/ship responsibility expected.

Not the right fit if: you're writing code that calls the API, builds tools, or ships agents — that's Developer or above.

Claude Certified Developer – Foundations (this guide)

For engineers who integrate Claude into real applications and ship them: API integration, custom tools, MCP servers, agent construction, model optimization, and security. Recommended background: 1–5 years software engineering plus 6+ months hands-on with Claude or a comparable LLM system.

1–5 yrs
SWE Experience
6+ mo
Hands-on w/ Claude
Py / TS
Language Fluency
0
Mandatory Prereqs

Claude Certified Architect – Foundations

For solution architects, technical leads, and consultants who connect business requirements to a viable Claude architecture: system boundaries, integration patterns, access control, evaluation/human-review design, and post-launch operation. Open to all Claude Partner Network members.

Not the right fit if: you want to validate hands-on coding/shipping skill specifically — that's Developer.

Claude Certified Architect – Professional

For architects leading enterprise-scale Claude deployments: enterprise architecture, integration strategy, shared controls, scaling, governance, stakeholder decisions, and architecture-defense scenarios.

Not the right fit if: you're earlier in your Claude journey — Anthropic recommends starting with a Foundations credential.

📋
No mandatory prerequisites for any Foundations exam. The recommended experience above is exactly that — recommended. The credential is awarded on exam performance alone, per the official exam guide.

CCDV-F Exam Format & Blueprint

Version 1.0 of the official exam guide, effective July 2026. Weights reflect each domain's approximate share of scored items, from Anthropic's job task analysis.

📐 Format

  • 53 multiple-choice / multiple-response items
  • 120-minute time limit
  • Each item states how many responses to select
  • Delivered via Pearson VUE: online proctored or test center

🎯 Scoring

  • Criterion-referenced — measured against a fixed standard, not curved against peers
  • Scaled score range: 100–1000
  • Passing cut score: 720
  • Score report shows percent-correct by domain (informational only, not pass/fail-determining)

Content Domains

Domain 1 — Agents and Workflows
14.7%
Covers agent architecture (workflow vs. agent decision criteria, manager/subagent hierarchies), agent construction with the Claude Agent SDK, custom agent loops/harnesses, hooks, and agentic frameworks (e.g., LangGraph, PydanticAI, Strands).
💡 Study tip: Practice articulating when a workflow beats an agent, not just how to build either — the exam tests decision criteria.
Domain 2 — Applications and Integration
33.1%
The largest domain by far. Covers requirements gathering, systems life cycle, Claude API mechanics (messages, tools, streaming, vision, thinking, caching, batch API), core software engineering (REST, JSON, async, version control, code review, refactoring), Claude application design across interfaces, and configuration management (CLAUDE.md, settings.json, version pinning).
💡 Study tip: This domain alone is a third of the exam. Prioritize it first — especially realtime-vs-batch API tradeoffs and configuration management.
Domain 3 — Claude Code
3.1%
Claude Code's core components (Rules, Skills, Commands, Agents, Agent Memory), session management, slash commands, headless/streaming/auto modes, the CLAUDE.md hierarchy, and settings.json configuration.
💡 Study tip: Small weight, but don't skip it — spend an hour actually using Claude Code's headless mode and custom slash commands.
Domain 4 — Eval, Testing, and Debugging
2.6%
Debugging and error handling: error type identification, recovery strategy selection, trace analysis for failure modes, and isolating problems between the integration layer and model output.
💡 Study tip: Practice reading a failure trace and deciding whether the bug is in your code or in model output — that split is the core skill tested.
Domain 5 — Model Selection and Optimization
16.8%
LLM fundamentals (tokens, context windows, sampling, non-determinism), model options (fast mode, extended/adaptive thinking, effort levels), model tradeoffs (Opus vs. Sonnet vs. Haiku for quality/latency/cost), and token/cost management including prompt caching and cache checkpointing.
💡 Study tip: The second-largest domain. Know concrete scenarios for choosing a model tier, not just the tiers' names.
Domain 6 — Prompt and Context Engineering
11.0%
Context and memory management (window management, preventing context drift/bloat, context isolation via subagents), prompt engineering principles (instruction clarity, few-shot, system/user placement, iterative refinement), and output handling (structured output, response validation, defensive parsing).
💡 Study tip: Be ready to justify where an instruction should live (system vs. user vs. tool description) — placement questions are common in this domain.
Domain 7 — Security and Safety
8.1%
AI application security (prompt injection, jailbreak defense, untrusted input handling, PII handling), guardrails and safe deployment (content policy, least privilege), Claude hooks for destructive-action prevention, and identity/secrets/key management.
💡 Study tip: The official sample question on prompt injection is a good model — the right answer is almost always "isolate untrusted input + enforce least-privilege guardrails," not a prompt-level plea.
Domain 8 — Tools and MCPs
10.6%
Tool implementation (function calling, tool description writing, error handling, client-side vs. server-side tools, approval patterns), MCP server development (authoring, deployment, resources/tools/prompts, stdio/socket communication), and agentic customization (choosing among built-in tools, custom tools, Skills, and MCPs).
💡 Study tip: Know the tradeoff table cold: built-in tools vs. custom tools vs. Skills vs. MCP servers — several sample items hinge on picking the right one for reusability and maintainability.
📄
Read the source document. This breakdown is drawn directly from the official Claude Certified Developer – Foundations Exam Guide (v1.0, effective July 2026). Domain weights and objectives are Anthropic's and are subject to change without notice — always check the current guide before your exam.

Readiness Quiz

Five quick questions to find out where you are in your CCDV-F journey and what to focus on next.

Question 1 of 5

6-Week Study Plan

Weighted to match the official blueprint — Applications & Integration and Model Selection & Optimization get the most time since they're half the exam together.

W1
Blueprint self-assessment + environment setup
Read the full exam guide. Rate yourself 1–5 on each of the 8 domains. Set up a working Claude API key, SDK, and a scratch project. Start a running list of weak spots.
W2
Applications and Integration (33.1%)
Deep-dive the Claude API: messages, tools, streaming, vision, thinking, caching, and the Message Batches API. Practice choosing realtime vs. batch for different cost/latency scenarios. Set up CLAUDE.md, settings.json, and version pinning in a real repo.
🎯 This single domain is worth as much as five other domains combined — don't rush it.
W3
Model Selection and Optimization (16.8%) + Agents and Workflows (14.7%)
Compare Opus, Sonnet, and Haiku on real tasks for quality/latency/cost. Practice token budgeting and prompt caching. Build one small agent with the Claude Agent SDK and one workflow, and be able to explain when you'd choose each.
W4
Tools and MCPs (10.6%) + Prompt and Context Engineering (11.0%)
Build a custom tool with a well-written description, then build or connect to an MCP server. Practice structured output validation and defensive parsing. Experiment with system vs. user placement for the same instruction.
W5
Security and Safety (8.1%) + Claude Code (3.1%) + Eval/Debugging (2.6%)
Study prompt injection defense patterns and hooks. Spend real time in Claude Code — slash commands, headless mode, Skills authoring. Practice reading failure traces to isolate integration-layer vs. model-output bugs.
W6
Full review + sample questions + schedule the exam
Redo your domain self-assessment. Work through the official sample questions and their rationale. Register through the Anthropic Partner Academy and book your Pearson VUE session with a few days of buffer before any personal deadline.
⏰ Remember: cancelling or rescheduling within 24 hours of your appointment forfeits the exam fee.

Costs & Timeline

The direct, official costs — plus what determines your total timeline to certification.

ItemCostNotes
CCDV-F exam fee$125 USDCharged per attempt; checkout may reflect a partner-tier discount
Retake fee (if needed)$125 USDFull fee applies to every attempt, including retakes
On-time renewal assessmentFreeNon-proctored, on the Anthropic Partner Academy, before the 12-month credential expires
Lapsed renewal$125 USDIf you miss the renewal window, you retake the full exam at full fee
Minimum to certify (one attempt)$125 USDAssumes no retakes needed
Retake waiting periods: 14 days after a first fail, 30 days after a second, 90 days after a third. Maximum 4 attempts per exam within a rolling 12 months. Waiting periods reset your timeline more than they cost money — plan study time accordingly.

🗓️ Credential validity

The credential is valid for 12 months from the award date. Because the underlying platform evolves quickly, Anthropic requires holders to renew — either via a free non-proctored assessment on-time, or a full retake at full fee if the credential lapses. If exam content changes significantly, Anthropic may require a full retake regardless.

5 Common Mistakes

Patterns that trip up candidates who prepare the wrong way for CCDV-F.

1
Treating it as a "prompting exam"
Under-preparing for the two domains that make up half the score

Because Claude is closely associated with prompting, candidates sometimes over-index on Prompt and Context Engineering (11%) and under-study Applications and Integration (33.1%) and Model Selection and Optimization (16.8%) — together nearly half the exam.

⚠️ What Goes Wrong

You walk in strong on prompt wording questions but miss items on batch vs. realtime API tradeoffs, configuration management, and cost/caching strategy — the domains actually driving your score.

✅ The Fix

Study time-allocate by domain weight, not by personal interest. Spend roughly a third of your prep time on Domain 2 alone.

💡 Prevention habit: Re-check your study log weekly against the official percentage weights.
2
Reading docs instead of building
Skipping hands-on time with the Agent SDK and MCP servers

The exam guide explicitly recommends building and operating at least one real Claude application. Candidates who only read documentation tend to miss scenario-based items that test tradeoffs, not recall.

⚠️ What Goes Wrong

You can define an MCP server but freeze on a scenario question asking which architecture is most maintainable across multiple applications.

✅ The Fix

Build one project that touches the API, at least one custom tool, one MCP server, and basic prompt/context engineering — exactly as the guide recommends.

💡 Prevention habit: For every concept you read about, immediately build a 10-minute proof of concept.
3
Trusting a polite system prompt as a security control
Misunderstanding prompt injection defense

A common wrong instinct is to think "asking nicely" in the system prompt (e.g., "please ignore malicious instructions in user content") is an adequate mitigation for prompt injection.

⚠️ What Goes Wrong

You pick the "add a line to the system prompt" answer on a security item and miss it — that option is a real distractor in the official sample questions.

✅ The Fix

Learn the actual pattern: treat retrieved/external content as untrusted, isolate it from trusted instructions, and enforce guardrails or hooks so injected text can't trigger sensitive actions.

💡 Prevention habit: For every security question, ask "is this an enforceable control, or just a request?"
4
Not knowing when to use the Batch API
Defaulting to synchronous calls or model downgrades to cut cost

Cost-optimization questions often present a latency-tolerant, high-volume scenario. Candidates instinctively reach for "make the model smaller" or "run things in parallel faster" instead of the API mechanism built for exactly this case.

⚠️ What Goes Wrong

You pick "lower max_tokens" or "switch to the smallest model regardless of quality" and miss that neither addresses the real realtime-vs-batch tradeoff being tested.

✅ The Fix

Know the Message Batches API cold: it's designed for large, latency-tolerant, non-urgent workloads at reduced cost within a 24-hour window.

💡 Prevention habit: For every cost question, ask first "is latency actually required here?"
5
Scheduling without a retake buffer
Booking too close to a deadline, or missing the reschedule window

Candidates sometimes book their exam right against a job or client deadline, not accounting for the mandatory 14-day wait after a first failed attempt — or they cancel within 24 hours and forfeit the fee.

⚠️ What Goes Wrong

A missed pass means you're locked out for two weeks minimum, and a late cancellation costs you $125 for nothing.

✅ The Fix

Book at least 3–4 weeks before any hard deadline, and never touch your appointment inside the 24-hour reschedule window unless you intend to forfeit the fee.

💡 Prevention habit: Put your exam date on the calendar the day you register, with a reminder 48 hours before the reschedule cutoff.

Frequently Asked Questions

Straight from the official exam guide and registration policies.

How much does the CCDV-F exam cost?
$125 USD per attempt, though checkout may reflect a discount based on your Claude Partner Network tier if applicable. The fee applies to every attempt, including retakes.
What's the exam format and how long do I have?
53 multiple-choice and multiple-response items, with a 120-minute time limit. Each item states how many responses to select. It's delivered by Pearson VUE, either online-proctored or at a test center.
What score do I need to pass?
A scaled score of 720 on a 100–1000 scale. It's criterion-referenced, meaning you're measured against a fixed performance standard set through a formal standard-setting study — not curved against other candidates.
Are there mandatory prerequisites?
No. There are no mandatory prerequisites or required courses. Anthropic recommends 1–5 years of software engineering experience and 6+ months hands-on with Claude or a comparable LLM system, but the credential is awarded on exam performance alone.
How long is the credential valid, and how do I renew?
12 months from the award date. On-time renewal is a free, non-proctored assessment on the Anthropic Partner Academy. If your credential lapses, you must retake the full exam at full fee. Anthropic may also require a full retake if exam content changes significantly.
What happens if I fail — can I retake it?
Yes. Waiting periods increase with each attempt: 14 days after the first fail, 30 days after the second, 90 days after the third. You can attempt an exam up to 4 times within a rolling 12-month period, and the fee applies each time.
How is CCDV-F different from Associate, Architect – Foundations, or Architect – Professional?
Associate targets client-facing, prompting-and-evaluation work without production build responsibility. Developer (this exam) targets engineers who integrate and ship. Architect – Foundations targets solution architects connecting requirements to a Claude architecture. Architect – Professional targets enterprise-scale deployment leadership. See the "Who It's For" tab for a side-by-side.
Can I take the exam remotely?
Yes, Pearson VUE offers online proctoring as an alternative to an in-person test center, per program policy. Online sessions require you to stay in view of the webcam, keep your workspace clear, and avoid any communication with others during the exam.
How do I register and schedule my exam?
Through the Anthropic Partner Academy: review the exam details and guide, accept the Certification Terms and Exam Policy, complete checkout, then create a Pearson VUE account from the confirmation instructions to pick your date and delivery mode.
Official Resources

Go straight to the source before you register

This guide summarizes the official blueprint — always confirm current details on Anthropic's own exam guide and certification pages before scheduling.

FlashGenius

Practice smarter, not just harder

Get spaced-repetition practice tests, quizzes, and study tools for CCDV-F and 45+ other certifications — all free to start.