Exam preparationClaude Certified Architect – FoundationsExam guide
Claude Certified Architect Foundations exam guide (CCAR-F)
What the CCAR-F exam covers and how it is scored: 60 questions in 120 minutes, a scaled pass mark of 720, and five weighted domains.
Agentic Architecture & Orchestration carries 27% of the exam, more than any other domain. In one of the guide's sample questions, the right answer is a programmatic check, because prompt-based fixes rely on probabilistic compliance.
01 What is on the CCAR-F exam, and how is it scored?
You have the exam booked and a stack of notes, but no clear picture of the test itself. Start with the format: 60 questions, 120 minutes, and a passing score of 720 on a scaled range of 100–1,000.
The CCAR-F exam at a glance
- questions
- 60
- minutes
- 120
- passing score
- 720
- exam price
- $125 USD
- question types
- Multiple-choice
- Multiple-response
Save the 'The CCAR-F exam at a glance' graphic as a PNG image PNG for LinkedIn or Pinterest
Plan your pace around 120 minutes for 60 questions.
02 Which domains carry the most weight?
CCAR-F domains and their weights
The whole exam: 60 questions
- 1Agentic Architecture & Orchestration27%about 16 questions
- 2Tool Design & MCP Integration18%about 11 questions
- 3Claude Code Configuration & Workflows20%about 12 questions
- 4Prompt Engineering & Structured Output20%about 12 questions
- 5Context Management & Reliability15%about 9 questions
Save the 'CCAR-F domains and their weights' graphic as a PNG image PNG for LinkedIn or Pinterest
Applied to 60 questions, Agentic Architecture & Orchestration at 27% is about 16 questions. Claude Code Configuration & Workflows and Prompt Engineering & Structured Output, at 20% each, are about 12 apiece.
Tool Design & MCP Integration at 18% is about 11 questions, and Context Management & Reliability at 15% is about 9.
Give agentic architecture the most time, but even the lightest domain is 15% of the exam.
03 What do the questions look like?
Expect scenarios. In one sample, an agent skips get_customer in 12% of cases and looks up orders by the customer's stated name, causing incorrect refunds.
The key is a programmatic prerequisite that blocks lookup_order and process_refund until get_customer returns a verified ID. A stronger system prompt or few-shot examples rely on probabilistic compliance, and a routing classifier fixes tool availability, not ordering.
When errors have financial consequences, pick deterministic enforcement over prompting.
04 How should you prepare?
MockCert has 5 untimed domain quizzes that explain each answer, and 5 mock exams of 60 questions timed at 120 minutes with the official domain weights. Each answer updates your estimated exam score by domain, so you can see which domain still needs work.
Mock Exam 1 is free with a free account, which makes it a good first check of your timing.
Learn by domain with quizzes, then rehearse under the real time limit.
Summary
- 60 questions in 120 minutes, with a scaled 720 to pass.
- Five domains, weighted from 27% down to 15%.
- When critical business logic needs a set tool order, enforce it in code, not prompts.
- Use quizzes to learn, then timed mock exams to rehearse.
Sources and editorial status
Research checked Nov 3, 2026.