CCA-F Practice Exam Anthropic Claude Certified Architect — Foundations  ·  Unofficial Study Tool

CCA-F Practice Exam

60 randomised questions · Harder than the real exam · Fresh set every time

🔓 Unofficial · AI-generated from public docs · No real exam questions

Have a promo code? Enter it to start without an API key.
Used only in this session — never stored or transmitted elsewhere.

What is the CCA-F Certification?

The CCA-F (Claude Certified Architect — Foundations) is Anthropic's official certification for developers and architects building production AI systems with Claude. The exam validates hands-on knowledge of the Claude API, agentic orchestration patterns, prompt engineering, and deployment best practices.

60 Questions
120 Minutes
720 Pass Score
4 / 5 Domains per Sitting

Each exam sitting covers 4 of the 5 domains chosen at random, weighted equally at 25% each. This practice tool mirrors that format — every session picks 4 domains and distributes questions evenly across them.

How This Tool Works

  • Questions are freshly generated by Claude AI each session from publicly available Anthropic documentation — no two sessions are the same.
  • Exam starts as soon as the first question is ready; the rest stream in the background so you never wait.
  • Practice mode shows explanations and doc links after each answer. Exam mode withholds all feedback until the end.
  • Results break down your score by domain so you know exactly where to focus your study.

Five Exam Domains

Questions cover the official CCA-F blueprint. Each practice session randomly selects 4 of these 5 domains, matching the real exam format.

D1 · 27%

Agentic Architecture & Orchestration

Agent SDK loop lifecycle, stop_reason values, single-agent vs multi-agent coordination, subagent spawning, HITL escalation patterns, task decomposition, and Agent SDK hook types.

D2 · 18%

Tool Design & MCP Integration

Tool descriptions as routing selectors, structured error response schema, the 18-tool degradation cliff, tool_choice modes, Model Context Protocol architecture, and MCP server configuration.

D3 · 20%

Claude Code Configuration & Workflows

CLAUDE.md hierarchy, slash commands, hooks system (PreToolUse/PostToolUse), plan mode activation, headless CI integration with --print, settings.json scope, and allowed tools configuration.

D4 · 20%

Prompt Engineering & Structured Output

Few-shot prompting, chain-of-thought, tool_use for reliable JSON output, validation-retry loops, attention engineering, extended thinking with budget_tokens, and system prompt design.

D5 · 15%

Context Management & Reliability

Prompt caching (cache_control, TTL, cost savings), lost-in-the-middle effect, RAG architecture, context window strategies, error propagation in multi-agent systems, and confidence calibration.

Frequently Asked Questions

Does this tool contain real CCA-F exam questions?

No. Every question is generated fresh by Claude AI from publicly available Anthropic documentation. No proprietary or leaked exam content is used or stored. The tool operates in full compliance with Anthropic's usage policies.

Do I need an Anthropic API key?

Yes — your API key is used to call Claude during generation and is never stored on the server. It exists only in your browser session. If you have a promo code, you can start without providing your own key.

How difficult are the questions?

Intentionally harder than the real exam. Questions are scenario-based with concrete constraints, field names, and code details designed to build deep understanding rather than surface recognition. Passing here is a strong signal you are ready.

What does the results screen show?

Your overall score (pass/fail at 70%), a per-domain breakdown table, and a full review of every missed question with the correct answer, explanation, and links to the relevant Anthropic documentation.

Rejoining the server...

Rejoin failed... trying again in seconds.

Failed to rejoin.
Please retry or reload the page.

The session has been paused by the server.

Failed to resume the session.
Please retry or reload the page.