Claude Certification Program · Exam code CCDV-F · Guide v1.0 (July 2026)
Claude Certified Developer – Foundations
Your exam: Tuesday 1 December 2026, 13:00
53 items120 min (≈2 min 15 s per item)pass at 720 on a 100–1,000 scale$125 per attemptvalid 12 monthsPearson VUE, online or test center
Where the points are
Weights come from the official blueprint. The “≈ items” column is 53 × weight: about how many of your questions will come from each domain. Half the exam is Applications & Integration plus Model Selection.
How the questions work
Scenario-based, decision-focused. Most items describe a situation and ask which approach best fits. Usually all four options are real mechanisms; three are used in the wrong place.
Read for the constraint. “Overnight”, “a user is waiting”, “reusable across apps”, “must never”, “steps are known”: one phrase usually decides the answer.
Multiple-response items say how many to select. There is no partial credit to count on, so pick exactly that many.
A prompt is not a control. When the question says “must always” or “must never”, look for code that runs outside the model: a hook, a permission, an allow-list, a human approval.
Simplest thing that works. Between a workflow and an agent, or a small and a big model, the exam rewards the cheaper option that meets the stated requirement.
About a third is ordinary engineering. REST semantics, idempotency, versioning, code review, rollouts and config management are tested with Claude as the dependency.
Score report: pass/fail, a scaled score and your percent correct per domain. Retakes after 14, 30, then 90 days; up to 4 attempts per 12 months.
Source: the official exam guide (v1.0, effective July 2026), found as a PDF on a course-hosting site rather than on anthropic.com. Its content matches what prep sites report. Product details (model names, defaults, prices) change fast; the exam targets the July 2026 state, so learn the reasons behind each choice rather than exact numbers.
Study guide, by weight
Each domain lists its blueprint skills, what to know cold, and a table of question cues with the usual answer and the trap answer.
Practice questions
Original questions written to the blueprint; not from the live exam. Study mode shows the answer after each item. Mock exam draws 53 items in blueprint proportions with a 120-minute timer and scores at the end.
Your record on this device
Plan to Tuesday 1 December, 13:00
About an hour on weekdays and two on weekend days. Domains are taken in order of weight, the build project runs alongside (the guide recommends having built a real Claude app), and the last week is for mock exams. The current week is highlighted.
23 Oct – 1 Nov · 33% of the exam
Applications & Integration
Messages API: statelessness, system prompt, content blocks, every stop_reason.
Streaming events and mid-stream errors; Batches; Files API; images and PDFs.
Bedrock, Vertex AI, Foundry: when and what differs.
REST and idempotency, async with concurrency limits, semver, code review, incremental refactoring.
Config: CLAUDE.md vs system prompt, settings.json, pinned models, prompt versioning, rollouts, monitoring.
Dometrain: Getting Started with Claude Code, the whole course (you'll use Claude Code for the build project; it also covers CLAUDE.md, permissions, model choice and cost).
Mon 23: go through mock 1 misses; reread the guide for every domain under 75%.
Thu 26: mock exam 2.
Fri 27 – Sat 28: your two weakest domains, guide plus Study mode.
Sun 29: mock exam 3. Target: 80% or more.
Still under 70%? Decide whether to reschedule before Mon 30 Nov, 13:00; after that the fee is lost.
30 Nov – 1 Dec · exam
Last day and exam day
Mon 30: light review only, the cue tables in each domain. Online exam: run the Pearson VUE system test on the computer and network you'll use. Prepare your ID. Early night.
Your Dometrain courses: Getting Started with Claude Code and Deep Dive: Claude Code. Strong for Claude Code, configuration, hooks, Skills, subagents and using MCP; they don't cover the API, batches, caching or building an MCP server.