Mock exam
60 questions · 120 minutes · pass mark 720 of 1000. Questions are drawn per domain in the blueprint proportions, so the section you are weakest in is the section that costs you the most.
Timed mock exam
60 questions in 120 minutes, drawn to the official blueprint weights. You can flag questions and come back; the timer submits automatically when it runs out.
| Domain | Weight | Questions drawn | Bank |
|---|---|---|---|
| Agentic Architecture & Orchestration | 27% | 16 | 72 |
| Tool Design & MCP Integration | 18% | 11 | 52 |
| Claude Code Configuration & Workflows | 20% | 12 | 65 |
| Prompt Engineering & Structured Output | 20% | 12 | 67 |
| Context Management & Reliability | 15% | 9 | 62 |
How this mock exam works
The real CCAR-F exam is 60 items in 120 minutes, closed book and proctored. Items are multiple-choice and multiple-response, and each one tells you how many answers to select — so “choose two” is information, not a hint. This mock matches that format exactly, including the clock.
Questions are drawn from a bank of 318, sampled per domain in the official blueprint proportions rather than evenly: 16 from Agentic Architecture & Orchestration (27%), 11 from Tool Design & MCP Integration (18%), 12 from Claude Code Configuration & Workflows (20%), 12 from Prompt Engineering & Structured Output (20%) and 9 from Context Management & Reliability (15%). That is the point of the exercise — the domain you are weakest in is the one that costs you the most, and an evenly sampled test would hide exactly that.
Twelve of the items in the bank are the official sample questions published in the exam guide, reproduced verbatim. The rest are written against the same task statements, and every option carries its own explanation, not just the correct one — the exam is a judgement test, so knowing why the other three answers fail is most of the skill.
A caveat about the score
The real exam reports a scaled score between 100 and 1000, passing at 720. Scaling adjusts for the difficulty of the particular items you were given, and Anthropic has not published the function it uses — so nobody outside the programme can reproduce it, this site included.
The number here is a linear approximation onto the same range. Treat it as a relative gauge, not a prediction: a 690 does not mean you would miss the pass mark by 30 points. The figure worth acting on is the per-domain breakdown underneath it, which tells you where the next hour of study belongs.