Claude Certified Architect – Foundations

About the Claude Certified Architect – Foundations (CCAR-F) exam

The exam at a glance

Questions

60 questions across four scenarios, multiple choice and multiple response

Time

120 minutes

Pass mark

720 on a scale of 100 to 1,000

Exam fee

About US$125

Valid for

12 months; renew before it lapses with Anthropic's free renewal assessment

Where you sit it

Pearson VUE, online proctored or at a test centre

Prerequisites

None, but you must work for a Claude Partner Network organisation to register

As published in the vendor’s exam guide at the time of writing. Fees vary by country and change; the vendor’s page is the source of truth.

What is on the exam, domain by domain

1. Agentic Architecture & Orchestration

   The largest domain. Workflows versus agents, orchestration patterns, subagents and delegation, and designing multi-step systems that stay controllable.

   27% of the paper

2. Tool Design & MCP Integration

   Designing tools Claude can use well, and integrating systems through the Model Context Protocol.

   18% of the paper

3. Claude Code Configuration & Workflows

   Configuring Claude Code for a team: project memory, settings and permissions, hooks, commands and automated use.

   20% of the paper

4. Prompt Engineering & Structured Output

   Prompts that produce reliable, structured, machine-readable output.

   20% of the paper

5. Context Management & Reliability

   Managing the context window, long-running work and failure modes so the system stays dependable.

   15% of the paper

How to prepare

1. 1.Read the exam guide's thirty task statements and turn each into a question you can answer from experience. The exam is built on them.
2. 2.Build one small agent with a tool and a subagent, and configure Claude Code for a real repository. Scenario questions are far easier once you have made those choices yourself.
3. 3.Read Anthropic's guidance on building effective agents and on context engineering; the architecture questions follow its reasoning.
4. 4.Practise reading a scenario first and asking what constraint decides the design: cost, latency, reliability or control.
5. 5.Take Anthropic's free prep course, then practise papers here until you clear your target twice running.

Sample questions

Three questions in the exam’s style, written by Square 1 against the published domains. They are ours, not the vendor's. The answer and the reasoning are under each one; the practice papers on this page draw from a bank written the same way, with an explanation on every question.

1. Question 1 · Agentic Architecture & Orchestration

   A task always runs the same three steps in the same order: extract fields from a document, validate them, then write a record. Which design fits best?

   1. AA fixed workflow that calls Claude for each step in sequence
   2. BAn autonomous agent that decides which steps to run and in what order
   3. CA team of subagents that each choose the next task for themselves
   4. DA single prompt that asks Claude to plan the steps on every run

   Answer: A. When the steps and their order are known in advance, a predefined workflow is simpler, cheaper and easier to test than an agent. Agents earn their cost when the path cannot be fixed ahead of time.

2. Question 2 · Prompt Engineering & Structured Output

   Downstream code parses Claude's reply as JSON and sometimes fails because extra prose appears around the object. What is the most reliable fix?

   1. AUse the platform's structured output support with a defined schema, and validate the result
   2. BAsk Claude more firmly in the prompt to reply with nothing except the JSON object
   3. CRaise the temperature so Claude is less likely to repeat the same formatting mistake
   4. DStrip every character before the first brace and after the last one in the reply

   Answer: A. Schema-backed structured output, followed by validation, is the dependable way to get parseable JSON. Firmer wording helps but does not guarantee it, and trimming characters hides the problem without fixing malformed output.

3. Question 3 · Context Management & Reliability

   A long-running agent session starts ignoring instructions given near the start as the conversation grows. Which approach best addresses this?

   1. ASummarise or compact earlier context and keep key instructions where they stay in view
   2. BRepeat the entire conversation history in every new request to reinforce it
   3. CSwitch to a smaller model so the session runs faster and ends sooner
   4. DRemove the system prompt so there is more room left for the conversation

   Answer: A. As context grows, important early instructions compete with everything after them. Compacting history and keeping key instructions prominent keeps them effective; resending everything makes the problem worse, and dropping the system prompt removes the instructions altogether.

Questions people ask before they book

Who can take the Claude Certified Architect – Foundations exam?

At the time of writing, only staff of organisations in the Claude Partner Network, registering with a partner email address. Anthropic has not announced a date for opening it to everyone.

What are the scenarios?

The exam draws four scenarios at random from a bank of six, and the questions sit inside them. Practise reading a situation and finding the constraint that decides the design.

Is it harder than the Developer exam?

It asks different things. The Developer exam is about building an integration correctly; the Architect exam is about choosing the right design and trade-offs for a situation.

How long is the certification valid?

Twelve months, renewable before it lapses with a free renewal assessment; after a lapse you sit the full exam again.

Are these the real exam questions?

No. Every question here is ours, written against Anthropic's public exam guide. Anthropic's exam content is confidential, and practising on leaked questions would break its exam policy.