Exam Guide โ Blueprint, Logistics, Strategy
The certification family (July 2026 lineup)
Four exams share the Claude Certified family. You're almost certainly here for one of the two Architect rows โ but it helps to see where they sit next to the others:
| Exam | Code | Audience | Q / Format | Fee | Key domain |
|---|---|---|---|---|---|
| Claude Certified Associate, Foundations | CCAO-F | Business users (ops, marketing, PM) | 60, traditional | $99 | Output evaluation (21%) |
| Claude Certified Developer, Foundations | CCDV-F | API/SDK builders | 53, traditional | $125 | Applications & Integration (33%) |
| Claude Certified Architect, Foundations | CCAR-F | Solution architects | 60, scenario-based | $125 | Agentic architecture (27%) |
| Claude Certified Architect, Professional | CCAR-P | Senior architects | 63, traditional | $175 | Integration (19%) |
All four: 120 minutes, pass at 720 / 1000, Pearson VUE↗ (online proctored โ a remote human/AI monitor watches via webcam โ or at a test center), valid 12 months with a free on-time renewal (non-proctored refresher on the Partner Academy↗).
Format footnote: traditional = standalone multiple-choice questions, each independent; scenario-based = questions grouped under a shared system description (see below). "ops, marketing, PM" = operations, marketing, and product-management staff; "API/SDK builders" = developers using the Claude API (the Messages endpoint) or its software development kit (SDK).
Logistics verified August 2026 against the official CCAR-F exam guide (downloadable from the certification page) and the certification FAQ↗. Always re-download the current guide before paying; fees, counts, and policies change. Details sourced from the partner-tenant FAQ are marked as such and may change without public notice.
CCAR-F blueprint (memorize the weights)
These five weights are the single most useful thing to burn into memory before exam day โ they tell you exactly where your study hours should go. Heaviest slice first:
| # | Domain | Weight | This guide |
|---|---|---|---|
| 1 | Agentic Architecture and Orchestration | 27% | Domain 1 |
| 2 | Claude Code Configuration and Workflows | 20% | Domain 2 |
| 3 | Prompt Engineering and Structured Output | 20% | Domain 3 |
| 4 | Tool Design and MCP Integration | 18% | Domain 4 |
| 5 | Context Management and Reliability | 15% | Domain 5 |
The scenario format โ what makes CCAR-F unique
Here's the twist that sets this exam apart. Your 60 questions don't stand alone โ they arrive grouped under 4 scenario contexts drawn randomly from a bank of 6. Each scenario is a one-paragraph system description, and its cluster of questions probes how you'd architect, configure, debug, and harden that one system. Good news: the six possible contexts are known ahead of time (full walkthroughs in Scenarios):
- Customer support agent on the Agent SDK with MCP tools and human escalation
- Claude Code for team code generation โ slash commands,
CLAUDE.md, shared conventions - Multi-agent research system โ a coordinator delegating to specialized subagents
- Developer productivity tooling โ Read, Write, Bash, Grep, Glob + MCP servers
- Claude Code in CI/CD โ automated review and PR feedback
- Structured data extraction โ validating output against JSON schemas
๐ฏ Because scenarios repeat across questions, misreading a scenario costs you a cluster of points, not one. Spend a full minute on each scenario paragraph before its first question. Identify: where's the model boundary, what are the tools, what's deterministic, where can it fail.
Registration (via the Anthropic Partner Academy↗)
โ ๏ธ Eligibility โ the gate before every other step (verified from the official FAQ, August 2026): certification is currently available only to people at Claude Partner Network organizations, and registration requires a company-domain email on the org's partner record โ personal addresses (gmail, etc.) will not work, which is exactly why the SSO shows "this account does not exist." The route for an unaffiliated builder: your company applies to the CPN (claude.com/form/cpn-partner-application↗); once approved, certify under the recognized company domain (adding/changing domains on a partner record takes 7โ10 business days via partner-support@anthropic.com). Two silver linings from the same FAQ: an exam registration, once made, stays valid for 5 years, and prep (the courses, the downloadable exam guide) is open to everyone today. One more clarification that resolves a common confusion: most "Anthropic certified" posts you see on LinkedIn are the free Academy course completion certificates (Claude 101, AI Fluency, etc. โ open to everyone, officially Anthropic-issued, LinkedIn-addable) โ a different artifact from these proctored, partner-gated exams. Note also that the CPN application page (as displayed August 2026) lists qualification guidance for the services network โ a registered business with 10 or more employees offering consulting/implementation/managed services โ and points product builders to the API/"Powered by Claude" track instead; application approval is Anthropic's discretion, and being Registered is distinct from qualifying for higher tiers.
๐ชค The two-tenant trap (author's experience, August 2026): there are TWO Skilljar instances that look identical. The public Academy (
anthropic.skilljar.com) hosts the free courses and takes ordinary email+password accounts. The certification lives on the Partner Academy tenant (anthropic-partners.skilljar.com), which signs in only through Anthropic's partner SSO โ and if that SSO has no partner identity for your email, the "we'll send a code" screen silently sends nothing, and Google sign-in reports "this account does not exist" (anti-enumeration: the site won't reveal whether an account exists). In my testing, no self-serve path created that identity โ access is provisioned through your organization's partner record. If you believe your org IS a partner and you still can't sign in, confirm you're using your company-domain address on the recognized domain, then contact partner-support@anthropic.com (domain changes take 7โ10 business days per the official FAQ). Your exam journey ultimately spans four account systems: public Academy (courses) โ partner SSO (exam registration) โ checkout โ Pearson VUE (delivery, keyed by the Candidate ID they email you โ and the Pearson name must exactly match your government ID).
Getting from zero to a scheduled seat is a short, well-marked path. Here it is, in order:
- Your organization joins the Claude Partner Network↗ (new applicants start at the Registered tier) โ which unlocks the Partner Academy↗ for its people.
- Open your certification's page on the partner tenant↗ โ download the official exam guide (do this before paying; it's the authoritative blueprint).
- Checkout ($125 / $175; partner discounts auto-apply).
- Create your Pearson VUE↗ account from the confirmation email.
- Schedule online-proctored or test-center. Cancel/reschedule up to 24 hours before.
Retakes: wait 14 days after a first fail, 30 after the second, 90 after the third; max four attempts per rolling 12 months; each attempt is full fee.
Renewal: on-time renewal is a free, non-proctored assessment covering what changed โ Anthropic keeps the window short (12 months) deliberately because the platform moves fast.
Official preparation: Anthropic Academy↗
Anthropic Academy↗ (launched March 2026) hosts free self-paced courses โ the official prep. Mapping to the architect track:
| Course | Feeds domains |
|---|---|
| Claude Code 101 ยท Claude Code in Action | 2 (Claude Code), 1 (agentic workflows) |
| Introduction to MCP ยท MCP: Advanced Topics | 4 (Tools & MCP) |
| Building with the Claude API | 3 (prompting/structured output), 5 (context) |
| Cloud-specific courses (Amazon Bedrock = AWS ยท Vertex AI = Google Cloud โ the cloud platforms that also serve Claude) | CCAR-P Integration |
A 3-day compressed study plan (exam in ~72 hours)
Cutting it close? This is the triage version: hit the heaviest domains first, and don't skip the redraw-from-memory step โ reproducing the diagrams by hand is what actually makes them stick.
| Day | Focus | Output |
|---|---|---|
| 1 | Domains 1 + 4 (45% combined): agent loop, orchestration patterns, tool schemas, MCP architecture | Redraw the agent-loop and MCP diagrams from memory |
| 2 | Domains 2 + 3 (40%): Claude Code config surface (CLAUDE.md, skills, hooks, settings, headless), prompting + structured output patterns |
Write a CLAUDE.md + one skill from scratch; do Samples |
| 3 | Domain 5 + Scenarios + Practice + Cram Sheet | 30 practice Qs; review misses; sleep |
Exam-day tactics
- 120 min / 60 Q = 2 min per question โ but scenario reading is spread across its cluster of questions: budget 1 min per scenario paragraph, ~90s per question, leaving a 10-minute review pass.
- Flag-and-move on any question fighting you; scenario clusters mean later questions often disambiguate earlier ones.
- Watch for the "most appropriate" stem: several answers work; the exam wants the one matching Anthropic's published guidance (simplest-thing-that-works, workflows-before-agents, deterministic-where-possible).
- Online proctoring: wired internet, clean desk, government ID, no second monitor, webcam room scan. Test center removes those variables โ worth it if your setup is uncertain.
๐ชค The classic architect-exam trap family: answers that are technically possible but violate a stated constraint in the scenario (cost ceiling, latency budget, no-human-in-loop, compliance boundary). Re-read the constraint sentence before choosing.