CCA-F Exam Guide: how to pass Claude Certified Architect — Foundations
The Claude Certified Architect — Foundations (CCA-F) is Anthropic's foundational certification for people who design and build with Claude. This guide breaks down what the exam covers, how it's scored, and the fastest way to prepare — including a free, interactive practice exam.
The exam validates that you can architect agentic systems with Claude — the Messages API, Agent SDK, Claude Code, and MCP — not just recall facts.
Can you actually sit this exam?
Worth checking before you invest study time. Anthropic describes the Claude Certified Architect, Foundations certification as earned through Partner Academy exams inside the Claude Partner Network, so access runs through a partner organisation rather than open public registration (Anthropic — Claude Partner Network, checked 2 September 2026).
Third-party guides also refer to the exam by the code CCAR-F and describe delivery through Pearson VUE. Neither is confirmed by Anthropic's own materials as far as we can establish, so treat both as unverified and check the official certification pages before booking anything.
If eligibility rules you out, the material still earns its keep. Two tracks on this platform need no eligibility at all and are open in full, free, without an account: the enterprise architecture reference and the Claude engineering curriculum.
Who the CCA-F is for
The target candidate is a solution architect with roughly six months of hands-on experience across the Claude API, the Agent SDK, Claude Code, and the Model Context Protocol (MCP). If you've built agentic loops, wired up tools, and shipped a Claude-powered workflow, you're in the right audience. The exam launched in 2026.
Exam format
- Length: roughly 60 scenario-based multiple-choice questions in a 120-minute session (confirm current specifics on Anthropic's official page). This trainer's own mock runs 90 minutes over 60 questions — deliberately tighter than the real thing, so pace is never what catches you out on the day.
- Question style: each question has one correct answer and three distractors — pick the single best option. There's no penalty for guessing, and unanswered questions count as incorrect.
- Scenario-based: the exam draws from 6 architecture scenarios, and 4 of the 6 are randomly selected each session. Questions sit inside realistic build-and-architect situations — agent design, multi-agent coordination, coding workflows, and data extraction — so you reason about trade-offs rather than recall trivia.
- Scoring: a scaled score from 100 to 1000; you need 720 to pass. Scaled scoring keeps difficulty fair across exam forms.
The five domains (and their weights)
Spend your study time in proportion to the weights. Agentic architecture alone is more than a quarter of the exam.
| # | Domain | Weight |
|---|---|---|
| 1 | Agentic Architecture & Orchestration | 27% |
| 2 | Tool Design & MCP Integration | 18% |
| 3 | Claude Code Configuration & Workflows | 20% |
| 4 | Prompt Engineering & Structured Output | 20% |
| 5 | Context Management & Reliability | 15% |
1 · Agentic Architecture & Orchestration (27%)
The heaviest domain. Agentic loops (continue on tool_use, stop on end_turn), coordinator–subagent patterns, isolated subagent context, enforcement vs. prompt guidance, hooks for interception, and session resumption/forking.
2 · Tool Design & MCP Integration (18%)
Designing clear tool interfaces and integrating MCP servers — descriptions that drive correct tool selection, structured error responses, and avoiding overlapping/ambiguous tools.
3 · Claude Code Configuration & Workflows (20%)
Configuring Claude Code: CLAUDE.md, custom slash commands, plan mode vs. direct edits, built-in tools, and using Claude Code in CI/CD for review, tests and PR feedback.
4 · Prompt Engineering & Structured Output (20%)
Reliable prompting and getting machine-readable output — schemas, validation, and high-accuracy extraction from unstructured text.
5 · Context Management & Reliability (15%)
Managing the context window, normalizing heterogeneous tool results, and building for reliability across long-running tasks.
How to prepare (a fast plan)
- Map the blueprint. Use the weights above to prioritize — agentic architecture first.
- Learn by doing. Reading isn't enough for a scenario exam. Build a small agentic loop, a custom tool, and an MCP integration.
- Drill with practice questions. Take full mock exams under the real format, then review every miss by domain.
- Close the gaps. Re-study the domains where your accuracy is lowest, then re-test.
Free interactive CCA-F practice exam
Lessons, cheat sheets, SM-2 flashcards and full mock exams — mapped 1:1 to the blueprint. Try the demo without signing up.
Start practicing free →Frequently asked questions
- What score do you need to pass the CCA-F?
- A scaled score of 720 on the 100–1000 scale. Scaled scoring converts your raw correct answers onto a 100–1000 scale that's equalized across exam forms.
- How many domains are on the exam?
- Five — agentic architecture (27%), tool design & MCP (18%), Claude Code (20%), prompt engineering & structured output (20%), and context management & reliability (15%).
- Do I need coding experience?
- Yes. The exam targets solution architects with about six months of hands-on Claude API, Agent SDK, Claude Code and MCP experience.
- Is the practice exam free?
- Yes — the trainer is free, bilingual (EN/UA), and the demo needs no sign-up.