Short answer
The Anthropic certification — also searched as the "Claude certification" — is the Claude Certified Architect – Foundations, officially coded CCAR-F and widely written CCA-F: Anthropic's official credential for engineers who build with Claude. It is a scenario-based exam across five domains — agentic architecture, tool/MCP design, Claude Code, prompt engineering, and context management — with a pass mark of
This page is the summary: what it is, the domains and weights, who should take it, the format and cost, how to prepare, and an honest read on whether it's worth it. For the full side-by-side exam blueprint, see the AI certification exam guide.
"Anthropic certification" and "Claude certification" are two names for the same thing: Anthropic's official credential program for people who build with Claude. The first exam in that program is the Claude Certified Architect – Foundations (CCAR-F), launched in March 2026. It validates that you can design and operate agentic systems on Claude — not that you can train models or recite trivia.
The exam carries "Architect" in its title, but the content is hands-on and implementation-focused. Anthropic positions it for software engineers, solutions architects, and AI developers — the people who actually wire Claude into production. If you build agents, tools, or Claude Code workflows, this is the credential that maps to your work.
The codes are the second source of confusion, and both are worth knowing. The official registration listing at Pearson VUE gives four codes — CCAR-F (Architect – Foundations), CCAR-P (Architect – Professional), CCAO-F (Associate – Foundations) and CCDV-F (Developer – Foundations). Many prep sites and forum threads instead write the Architect – Foundations exam as CCA-F, and some split the difference with "CCA-F/CCAR-F". They are the same exam. CCAR-F is the code Anthropic and Pearson VUE use, so it is the one to look for once you have Partner Network access; CCA-F is the one you are more likely to meet in third-party prep material.
One disambiguation up front
The Anthropic / Claude certification (CCAR-F) isnotthe AWS GenAI exams. AWS has its own credentials —AIP-C01(Generative AI Developer – Professional) and the foundationalAIF-C01(AI Practitioner) — which test Bedrock, not Claude specifically. The full split is in the[exam guide], and the CCAR-F sits next to the AWS, Google and Microsoft exams in the[AI certifications compared]table if you want to see the whole field at once.
The exam is built from five weighted domains. Map your study time to the weights — the blueprint is the syllabus. The domain names and weights below are confirmed against Anthropic's official exam guide.
| Domain | Weight |
|---|---|
| D1 — Agentic Architecture & Orchestration | 27% |
| D2 — Tool Design & MCP Integration | 18% |
| D3 — Claude Code Configuration & Workflows | 20% |
| D4 — Prompt Engineering & Structured Output | 20% |
| D5 — Context Management & Reliability | 15% |
The shape of the exam is clear from the weights: D1 Agentic Architecture & Orchestration (27%) is the single heaviest domain, with D3 and D4 tied at 20% just behind. A blind spot in agentic design costs you far more marks than one in D5. For the per-domain task statements and the AWS exams side by side, the exam guide has the detail; this page keeps it to the summary.
The CCAR-F fits engineers who already build software and want a Claude-specific signal on top:
It is a Foundations exam, so it assumes general engineering maturity rather than years of LLM-specific work. If you're weighing it against the AWS path, the role it maps to is the AI architect — and your model-selection judgement is exactly the kind of reasoning the scenarios test.
Anthropic can change exam details. Pass mark, scenario format, domain weights, question count, time limit, and fee are all confirmed against the official v1.0 exam guide (2026-07-08) — verify on Anthropic's certification page before relying on them.
Because the exam is scenario-based, prep is about reasoning through realistic situations, not memorizing definitions. A pragmatic path:
That's how aiArch's curriculum is mapped — its modules track the CCAR-F domains, so your study time lands where the marks are.
The provenance goes below the domain level. Every practice item in the aiArch item bank carries its own source list, shown under the answer once you've committed to one, and every lesson carries its own source list behind the same gate — a content gate refuses any item or lesson without sources, so none ships without them. You can check that without an account: the free sample lesson shows a source line under each of its questions.
These are original sample questions we wrote — grounded in the CCAR-F blueprint and Anthropic's public documentation, not taken from any real exam. Each answer shows the reasoning and cites its source. The five questions are an interactive widget on the original page — take them at aiarch.dev/anthropic-certification#quiz — so they do not survive syndication.
The knowledge the CCAR-F tests is worth it regardless — it maps directly to real agentic engineering skills you'll use whether or not you sit the exam. The credential itself is partner-gated, so the badge may be harder to obtain than the prep, which is open to everyone. For most seniors the move is: prep now for the knowledge, obtain the badge when you have Partner Network access.
As a hiring signal, a cert helps you pass filters — especially at consultancies and cloud partners — but it doesn't qualify you on its own; portfolio and interview performance weigh more. For the full ROI verdict, including how the CCAR-F compares to the AWS exams for your destination role, see which AI certifications are worth it. This page is the explainer; that one is the verdict.
The "how to prepare" advice above isn't generic study tips — it's the same mapping we used to build aiArch's own curriculum. docs/CURRICULUM.md
assigns Track C's modules to the five CCAR-F domains one at a time: D1 Agentic Architecture & Orchestration (27%) gets the deepest coverage across Tracks B and C, because that's where a quarter of the exam's marks sit, down to D5 Context Management & Reliability (15%), the lightest.
We hold this mapping to the same bar we apply to our own coach: don't claim more than we've verified. Every domain weight above is checked against Anthropic's official exam guide before it's used to justify a module's depth — a rule that came out of an earlier miss, where a curriculum revision quoted a domain split from a secondhand summary and had to be walked back once the primary guide confirmed different figures. See Anthropic's own materials and Claude's tool-use docs — the same API our production coach (src/lib/coach.ts
) calls in the bounded agentic loop that the D1 domain actually tests you on.
The Anthropic certification is Anthropic's official credential for people who build with Claude. Its first exam is the Claude Certified Architect – Foundations (CCAR-F), a scenario-based exam covering agentic architecture, tool and MCP design, Claude Code, prompt engineering, and context management, with a pass mark of 720 out of 1000.
Yes — "Claude certification" is just another name for the Anthropic certification. The official exam is the Claude Certified Architect – Foundations (CCAR-F), launched in March 2026 and aimed at software engineers, solutions architects, and AI developers who build on Claude.
CCAR-F stands for Claude Certified Architect – Foundations, Anthropic's foundational Claude credential. Despite the "Architect" label it is implementation-heavy. The exam is scenario-based, with multiple-choice and multiple-response items — each item states how many responses to select — and you sit 4 scenarios drawn from a bank of 6.
$125, per the official exam guide. The CCAR-F is partner-gated — obtained through the Claude Partner Network — but the fee itself is a confirmed published figure since Anthropic's 2026-07-08 four-exam certification portfolio launch.
Five weighted domains: D1 Agentic Architecture & Orchestration (27%), D2 Tool Design & MCP Integration (18%), D3 Claude Code Configuration & Workflows (20%), D4 Prompt Engineering & Structured Output (20%), and D5 Context Management & Reliability (15%). The weights are confirmed against Anthropic's official exam guide. For the per-domain detail, see the AI certification exam guide.
The knowledge is worth it regardless — it maps to real agentic engineering skills. The credential itself is partner-gated, so the badge may be harder to obtain than the prep. As a hiring signal it helps pass filters but doesn't qualify you alone. For the full ROI take, see our verdict on which AI certifications are worth it.
Map study time to the domain weights (D1 is heaviest at 27%), build a real tool-calling agent rather than a notebook demo, practice the Claude model-selection decision the scenarios reward, and use Anthropic Academy's free Claude courses alongside hands-on building. The exam is scenario-based, so reason through situations rather than memorizing facts.
Pass mark, format, domain weights, question count, time limit, and fee are confirmed against the official v1.0 exam guide (2026-07-08) — verify on Anthropic's certification page before you register. Corrections: hello@aiarch.dev.
Originally published at aiarch.dev/anthropic-certification, where it is kept up to date.
Want the skeleton instead of the essay? aiarch-templates has the src/lib/ seams, a threshold-gated eval stub and a cost-model skeleton. It is deliberately empty — it fixes the shape and you write the implementation. Apache-2.0.