Claude Certified Architect Foundations vs Professional (CCAR-F vs CCAR-P)
Anthropic's architect certification has two levels: Claude Certified Architect – Foundations (CCAR-F) and Claude Certified Architect – Professional (CCAR-P). They share a scoring model and differ in almost everything else — who they are for, what they weight, and what a question asks you to decide.
The two exams side by side
| Foundations (CCAR-F) | Professional (CCAR-P) | |
|---|---|---|
| Questions | 60 | 63 |
| Time | 120 minutes | 120 minutes |
| Time per question | two minutes | about one minute fifty |
| Pass mark | 720 on 100–1,000, total score only | 720 on 100–1,000, total score only |
| Registration | $125 | $175 |
| Answer format | one correct response, three distractors | one intended answer per item |
| Domains | 5 | 7 |
| Largest domain | Agentic Architecture & Orchestration, 27% | Integration, 19% |
| Attempts | four in a rolling 12 months | four in a rolling 12 months, with waits of 14, 30 and 90 days |
| Validity | 12 months, free renewal before it lapses | 12 months |
On both, 720 is a scaled score rather than a percentage, per-domain percentages are informational rather than separate bars, unanswered questions are scored as incorrect and there is no penalty for guessing.
Who each is for
Foundations is for solution architects building production applications on Claude. Alloy's Foundations course assumes you have prompted Claude and read its output — not that you have shipped an agent, written an MCP server or sat a certification before. The exam covers Claude Code, the Claude Agent SDK, the Claude API and the Model Context Protocol.
Professional is for mid-level and senior architects — typically three or more years of architecture experience and six or more months running Claude in production — who take a business problem to a scalable solution across its lifecycle. Alloy's Professional course assumes you already understand agent control flow and tool use.
What each weights
Foundations — Agentic Architecture & Orchestration (27%), Claude Code Configuration & Workflows (20%), Prompt Engineering & Structured Output (20%), Tool Design & MCP Integration (18%) and Context Management & Reliability (15%). Its centre of gravity is building: stop_reason-driven loops, subagent routing, plan–edit–verify, parseable output, and tool schemas the model uses correctly.
Professional — Integration (19%), Solution Design & Architecture (17%), Evaluation, Testing & Optimization (16%), Stakeholder Communication & Lifecycle Management (14%), Governance, Safety & Risk Management (14%), Claude Models, Prompting & Context Engineering (13%) and Developer Productivity & Operational Enablement (7%). Its weight is spread more evenly and reaches into the business: value pillars, evaluation design, compliance, stakeholder discovery and the lifecycle after handoff.
How the questions differ
Foundations questions are scenarios drawn from Anthropic's official scenarios — including Customer Support Resolution Agent, Code Generation with Claude Code, Multi-Agent Research System, Developer Productivity with Claude, Claude Code for CI/CD and Structured Data Extraction — with four presented on each exam form. Its published out-of-scope list rules out topics such as fine-tuning, computer use, vision, streaming, tokenization and MCP server hosting.
Professional questions typically give a business situation, a constraint that narrows it, and four architectures that all look reasonable. The constraint — a latency budget, data residency, a human-review requirement, a cost ceiling, an existing system to integrate with — decides. Options are often individually defensible and contextually wrong.
On both, the same five wrong-answer shapes recur: the absolutist, true but not the question, right action at the wrong layer, the over-engineered option, and the answer to a different question. And on both, change an answer only when you can name what you missed.
Which to take first
Neither exam has a prerequisite — none of Anthropic's four Claude certifications does — so you can sit Professional without holding Foundations. If you are building your first production agents, Foundations is the on-ramp: its heaviest domain is the control flow the Professional course takes for granted. If you already design and run Claude systems in production, and your work includes justifying designs to the business, Professional tests that directly. The free diagnostics are the quickest way to decide: take both, 24 questions each, about 20 minutes, and compare where you stand.
How Alloy prepares you for each
Foundations — thirteen lessons: an exam-strategy lesson and twelve domain lessons (three each for agentic architecture and Claude Code workflows, two each for the other three domains), with hands-on labs where you build a real loop.
Professional — fifteen lessons: an exam-strategy lesson and fourteen domain lessons (three for Integration, two for each of five domains, one for Developer Productivity), with scenarios built around constrained architecture choices.
On both courses the first two lessons are free with their podcasts: the exam-strategy lesson, and a first domain lesson — agentic loops and stop_reason control flow for Foundations; splitting a request among Claude, existing systems and people and pricing it for Professional. Alloy's Full membership covers the prep for both exams, and for the other two Claude certifications; the exam registrations are separate. Prices are on the pricing page.
Next step
Create a free account — no card needed — and take the diagnostic for either exam, or both.
Alloy is exam preparation built from the official exam guides — only what the exam asks for.