Start studying freeStart free

Anthropic · Foundations

CCAR-F exam prep

Claude Certified Architect – Foundations

Design with Claude: agentic loops and multi-agent systems, tools and MCP, Claude Code, structured output and reliability. The exam is open to Claude Partner Network members.

Start studying CCAR-F free Sample questions
practice questions
473
chapters
41
slide decks
34
practice exams
2

The exam at a glance

Items
60, multiple-choice and multiple-response
Structure
4 scenarios drawn from a bank of 6
Time
120 minutes
Passing score
Scaled 720 on 100–1,000
Validity
12 months from award

Source: Anthropic, Claude Certified Architect – Foundations Exam Guide v1.0 (effective July 2026)

What the exam covers

Every objective in the official study guide has a chapter, a slide deck and practice questions.

  1. Agentic Architecture & Orchestration

    27% of the exam

    1. 1.1 Agentic loops
    2. 1.2 Coordinator-subagent orchestration
    3. 1.3 Subagent invocation and context passing
    4. 1.4 Workflow enforcement and handoffs
    5. 1.5 Agent SDK hooks for interception and normalization
    6. 1.6 Task decomposition strategies
    7. 1.7 Sessions, resumption, and forking

  2. Tool Design & MCP Integration

    18% of the exam

    1. 2.1 Tool interface design
    2. 2.2 Structured error responses for MCP tools
    3. 2.3 Tool distribution and tool choice
    4. 2.4 MCP servers in Claude Code and agent workflows
    5. 2.5 Built-in tools: Read, Write, Edit, Bash, Grep, Glob

  3. Claude Code Configuration & Workflows

    20% of the exam

    1. 3.1 CLAUDE.md hierarchy and modular configuration
    2. 3.2 Custom slash commands and skills
    3. 3.3 Path-specific rules
    4. 3.4 Plan mode vs direct execution
    5. 3.5 Iterative refinement
    6. 3.6 Claude Code in CI/CD pipelines

  4. Prompt Engineering & Structured Output

    20% of the exam

    1. 4.1 Explicit criteria and false positives
    2. 4.2 Few-shot prompting
    3. 4.3 Structured output with tool use and JSON schemas
    4. 4.4 Validation, retry, and feedback loops
    5. 4.5 Batch processing strategies
    6. 4.6 Multi-instance and multi-pass review

  5. Context Management & Reliability

    15% of the exam

    1. 5.1 Preserving context in long interactions
    2. 5.2 Escalation and ambiguity resolution
    3. 5.3 Error propagation in multi-agent systems
    4. 5.4 Context in large codebase exploration
    5. 5.5 Human review and confidence calibration
    6. 5.6 Provenance and uncertainty in synthesis

  6. Foundations

    Background, not scored

    1. 0.1 The Claude platform for architects
    2. 0.2 How Claude behaves in production
    3. 0.3 Claude Code and the Agent SDK
    4. 0.4 MCP essentials

Sample questions

Three original questions from the bank, with their explanations.

Agentic Architecture & Orchestration · From chapter 1.1 Agentic loops

Customers hear "Let me check" and nothing more.

Harbor & Pine's support agent runs on a custom Messages API loop. In about 15% of conversations, the agent's last message is "Let me pull up that order for you," and the tool logs show no order lookup. The loop ends the run as soon as a reply contains a text block.

What is the most likely root cause?

Choose an answer

Claude Code Configuration & Workflows · From chapter 3.1 CLAUDE.md hierarchy and modular configuration

A money rule works on one laptop but not on dev boxes.

Corvid Bank's tech lead wrote "store amounts as integer minor units, never floats" months ago. Claude follows it on her laptop, but on the team's shared cloud dev boxes it writes float arithmetic in the ledger. /context on a dev box lists only the repository's CLAUDE.md under Memory files.

What should you do?

Choose an answer

Prompt Engineering & Structured Output · From chapter 4.1 Explicit criteria and false positives

Developers ignore the review bot, including its security findings.

Northgate's bot posts about 70 comments per 100 pull requests. A labeled sample shows security findings at 90% precision, correctness at 70%, comment accuracy at 20% and style at 10%. Developers now resolve bot threads unread, and a flagged injection bug shipped.

What should the team do first?

Choose an answer

How you will study

Learn each objective with a chapter and a slide deck, practice with a drill that brings back what you miss, then rehearse with timed exams weighted like the real one.

See how it works →

  1. 1

    Learn

    Understand it, then remember it

    Chapters with decision tables and exam traps · Slide decks with step-by-step animated diagrams · A study map of the whole exam on one page

  2. 2

    Practice

    Practice what you are about to forget

    Reviews spaced at 1, 3, 7, 14 and 30 days · Only the chapters you have finished, unless you choose more · An explanation for every option, right or wrong

  3. 3

    Pass

    Rehearse the real thing

    From a 5-question warm-up to a full-length exam · Longer Azure exams end with a complete case study · Mastery per domain against the weights the exam uses

FAQ

CCAR-F FAQ

What does the CCAR-F exam cover?

CCAR-F (Claude Certified Architect – Foundations) covers Agentic Architecture & Orchestration (27%), Tool Design & MCP Integration (18%), Claude Code Configuration & Workflows (20%), Prompt Engineering & Structured Output (20%) and Context Management & Reliability (15%). 3jumps teaches each objective in 41 chapters.

How many CCAR-F practice questions are there?

473 original practice questions, including 2 full practice exams. Every option has its own explanation.

How long is the CCAR-F exam and what is the passing score?

Time: 120 minutes. Passing score: Scaled 720 on 100–1,000. Confirm on the official exam page before you book.

Are these real CCAR-F exam questions?

No. Every question is original, written from the official objectives and checked against Anthropic's documentation. 3jumps never uses exam dumps and is not affiliated with Anthropic.

Start studying CCAR-F free

Open the first chapter and let the drill take it from there.

Start studying free