Claude Certified Architect, Foundations

10 tests · 100 questions
Browse all 100 questions ›
Your progress

Full practice exams

Practice tests

Study guides

About the CCA-F exam

Claude Certified Architect, Foundations (exam code CCAR-F) is Anthropic's certification for people who design and build production applications on Claude. It covers four technologies together: the Claude API, the Claude Agent SDK, Claude Code and the Model Context Protocol (MCP). Anthropic suggests around six months of hands-on experience building with them before you sit it.

The exam is about judgement more than recall. Questions are grouped under realistic scenarios, such as a customer support agent, a multi-agent research system or a document extraction pipeline, and ask what you would do next. The wrong options are usually things that look reasonable but fail in production: relying on a prompt instruction where a hook should enforce a rule, stopping an agent loop on a fixed iteration count, or throwing a bigger context window at a problem that needs better decomposition.

Exam format

  • Questions: 60, multiple choice and multiple response. Each question says how many answers to pick.
  • Structure: each sitting draws 4 scenarios at random from a published set of 6, with a group of questions under each.
  • Time: 120 minutes.
  • Passing score: 720 on a scaled score from 100 to 1,000. The score report also shows your percentage correct in each domain.
  • Cost: 125 USD, before any partner-tier discount. Registration is through the Anthropic Partner Academy.
  • Delivery: proctored by Pearson VUE, online or at a test center.
  • Who can take it: people who work at an organization in the Claude Partner Network, registering with a company email on that organization's domain, aged 18 or older with a government-issued ID.
  • Validity: 12 months. On-time renewal is a free, unproctored assessment; a lapsed credential means taking the full exam again.

What the exam covers

The exam guide (version 1.0, effective July 2026) has five domains:

  • Agentic Architecture and Orchestration (27%). The agentic loop driven by stop_reason, coordinator and subagent designs, passing context to subagents explicitly, hooks that intercept tool calls, task decomposition, and resuming or forking sessions.
  • Claude Code Configuration and Workflows (20%). CLAUDE.md hierarchy and scoping, custom slash commands and skills, path-specific rules, when to use plan mode, iterative refinement, and running Claude Code in CI/CD pipelines.
  • Prompt Engineering and Structured Output (20%). Explicit criteria to cut false positives, few-shot examples, structured output through tool use and JSON schemas, validation and retry loops, batch processing, and multi-pass review.
  • Tool Design and MCP Integration (18%). Writing clear tool descriptions, structured error responses from MCP tools, giving each agent the right tools, connecting MCP servers, and choosing between built-in tools such as Read, Grep and Bash.
  • Context Management and Reliability (15%). Keeping key facts across long conversations, escalation and human review, propagating errors in multi-agent systems, exploring large codebases, and keeping track of sources and uncertainty.

How the practice questions here are organised

There are 10 practice tests of 10 questions each, and every test covers all five domains, so any one of them is a fair cross-section rather than a single-topic quiz. Each question explains why the right answer is right and why each other option is wrong, with a link to Anthropic's documentation.

There are also two full practice exams that copy the real format: 60 questions in 120 minutes, weighted by domain as in the guide, with answers hidden until you submit. Each uses four of the six official scenarios, and between them all six are covered. Because this site has no separate scenario screen, every question restates the context it needs. Anthropic doesn't publish the raw cut score, so the mock passes you at 72%.

A study plan that works

  • Read the official exam guide in full first, including its sample questions. It lists every task statement the exam is written against.
  • Take Test 1 without preparing, then read every explanation, including for questions you got right by guessing.
  • Build something small with the Agent SDK: a loop with two tools, then a coordinator that hands work to a subagent. Many questions are easy once you have seen how context really flows.
  • Set up Claude Code on a real repository with a CLAUDE.md, a custom command and a hook, and try plan mode on a multi-file change.
  • Work through the rest of the tests, revisiting any domain where you score under 70%.
  • Finish with both full practice exams, timed and in one sitting each.

These questions are original, written from Anthropic's public documentation and the official exam guide. They are not real exam questions, and this site is not affiliated with Anthropic.