What the exam covers
CCAR-P reports on a scaled range of 100–1000 with a cut at
720, and its questions are distributed across
7 domains by published weight rather than evenly. Those
weights are the single most useful thing in the exam guide: they tell you
where the marks are, and they rarely match where candidates spend their time.
| Domain | Share of exam | Approx. questions |
|---|
| Domain 1 — Solution Design and Architecture |
17.0% |
11 |
| Domain 2 — Claude Models, Prompting, and Context Engineering |
13.0% |
8 |
| Domain 3 — Integration |
19.0% |
12 |
| Domain 4 — Evaluation, Testing, and Optimization |
16.0% |
10 |
| Domain 5 — Governance, Safety, and Risk Management |
14.0% |
9 |
| Domain 6 — Stakeholder Communication and Lifecycle Management |
14.0% |
9 |
| Domain 7 — Developer Productivity and Operational Enablement |
7.0% |
4 |
See all 20 objectives and their individual weights →
Integration is the heaviest at
19.0% — roughly
12 of the
63 questions. Weakness there costs more than weakness
anywhere else, which is exactly what a flat percentage score hides and what
a blueprint-weighted estimate surfaces.
Where to start
Three objectives carry the most weight on CCAR-P. Between them they
are worth roughly
12 of the 63 questions:
-
Tool and function design —
6.5%, about
4 questions
-
MCP servers: scope, auth and discovery —
6.5%, about
4 questions
-
Integrating with systems of record —
6.0%, about
4 questions
Which is not the same list as the objectives that feel hardest, and that
gap is where study time usually goes missing. A week spent on the topic
that worries you most is a week spent on
6% of the exam at best,
and often on far less.
How we practice it
Our CCAR-P bank holds 380 original items across
20 objectives, each with an explanation and a note on
every option — why the right answer is right, and why each distractor
fails. Items are written against the published objectives. None is recalled
or reproduced from a live exam.
A drill allocates questions by
the published domain weights,
multiplied by how far your measured mastery sits below standard.
So a domain
worth 19% of the
exam that you are weak in dominates your next sitting, while one you have
already mastered drops back to an exposure floor that keeps its estimate
current.
What a question looks like
Taken from the free CCAR-P set, not written for this page —
the correct answer, why it is correct, and a note on why each other option
fails. Scenarios, not recall: nothing here is answerable from a definition.
CCAR-P
Managing the context budget
Which context content should be evicted first when a long-running agent approaches its limit?
-
A
The system prompt, which has already shaped everything that was done so far
The system prompt governs behavior throughout and is the worst thing to drop.
-
B
The current task description, which the agent has been working from
The current task is what the agent is working on.
-
C
The most recent exchange, since it is the least embedded in the work
Recent exchanges are the immediate working state.
-
D
Raw tool output whose conclusions have already been drawn and recorded
Correct. The value has already been extracted, so the raw form is redundant.
Raw output that has already been reduced to a conclusion is the cheapest thing to lose: the value has been extracted. The system prompt, the current task and recent exchanges are all load-bearing for what happens next.
See all 20 free CCAR-P questions →
What you get for free
A 20-item diagnostic for CCAR-P, spread across
objectives rather than taken off the top, so the sample reflects the shape
of the exam. It produces a first readiness estimate and full explanations,
whether or not you go on to buy.
The other Claude certification exams
All four report on the same 100–1000 scale with the same
720 cut, so a readiness estimate means the same thing
whichever you are sitting. What differs is shape: