Claude Certified Architect: Foundations (CCAR-F)
226 practice questions, including 19 multiple-response scenario items in the format the exam actually uses. Every answer sourced from official Anthropic documentation. Built by a PharmD who has designed real certification exam questions before, and ships production software with Claude Code.
One-time purchase. Instant access.
Get Instant Access6 things most CCAR-F prep resources get wrong
Every fact in this bank is checked against the official Anthropic documentation. These six show up wrong across most competing study material, and each one is a real point lost on exam day.
Wrong CLAUDE.md project path
Most resources say project memory lives at .claude/CLAUDE.md. The real path is ./CLAUDE.md, in the project root. The .claude/ folder is for settings, rules, and commands, not CLAUDE.md.
SSE listed as a current MCP transport
SSE was deprecated in the MCP spec. The current transports are stdio and Streamable HTTP. Any resource still recommending SSE for new work is teaching you to fail a transport question.
Only 3 tool_choice values taught
There are four: auto, any, forced ({type:"tool",name:"X"}), and none. Missing none costs you on any text-only-turn question.
Hooks configured in CLAUDE.md
Hooks belong in settings.json, either ~/.claude/settings.json (user) or .claude/settings.json (project). CLAUDE.md is natural-language context; settings.json is parsed configuration.
Only 2 stop_reason values taught
There are three: tool_use, end_turn, and pause_turn. When pause_turn fires, the iteration limit was hit mid-work, Claude is not finished, and you must re-send.
One prompt caching threshold for every model
The minimum is 1,024 tokens for Opus and Sonnet, but 2,048 tokens for Haiku 3.5. The blanket "1,024 rule" that circulates everywhere is only correct for the Opus/Sonnet tier.
Sample Questions
1. Your agentic loop code checks: if (response.content[0].type === 'text') { break; }. In production, it sometimes stops mid-task. What is the root cause?
Claude can legitimately return both text content and tool_use blocks in the same response. Checking content[0].type == 'text' breaks the loop prematurely, abandoning the pending tool calls. Always check stop_reason.
2. A new developer cloned the repo but Claude Code doesn't follow the team's coding standards. The lead developer has the standards defined in ~/.claude/CLAUDE.md. What is the problem?
User-level config (~/.claude/CLAUDE.md) is personal, it lives outside the repository and is never shared via git. Team-wide conventions must be in project root ./CLAUDE.md, committed to the repository.
3. Claude keeps fabricating values for a 'tax_id' field when extracting from documents that don't contain tax IDs. What schema change prevents this?
Making fields optional/nullable signals to Claude that the field may legitimately be absent. When a field is required, Claude is forced to provide something, leading to fabrication.
223 more questions like these, organized by domain, with full explanations.
Get Full Access for $37All Five CCAR-F Domains Covered
What is on the CCAR-F exam?
This section describes Anthropic's exam itself, not this study pack. The Claude Certified Architect: Foundations exam is 60 multiple choice and scenario questions in 120 minutes. Anthropic reports a scaled score from 100 to 1,000 and sets the pass mark at 720. The exam costs $125, is delivered online proctored or at a Pearson VUE test centre, and the credential is valid for 12 months.
| Exam name | Claude Certified Architect: Foundations |
|---|---|
| Exam code | CCAR-F |
| Questions on the exam | 60, including scenario based items. This pack drills 226 |
| Time limit | 120 minutes |
| Passing score | 720 on a 100 to 1,000 scale |
| Exam fee, paid to Anthropic | $125. This study pack is $37 and is a separate purchase |
| Delivery | Online proctored, or a Pearson VUE test centre |
| Credential validity | 12 months |
| Domain weights | D1 27%, D3 20%, D4 20%, D2 18%, D5 15% |
Exam structure is 4 scenarios drawn from a bank of 6, and items are a mix of multiple choice and multiple response, with each item stating how many answers to select. If you do not pass, the waiting period is 14 days after a first attempt, 30 days after a second, and 90 days after a third, with a maximum of four attempts in a rolling twelve months. The fee applies to each attempt.
What is not on the CCAR-F exam?
The exam guide publishes an out of scope list, and it is worth reading before you over prepare. These do not appear on the exam:
- Fine tuning Claude models or training custom models
- Claude API authentication, billing, or account management
- Deploying or hosting MCP servers: infrastructure, networking, container orchestration
- Claude’s internal architecture, training process, or model weights
- Constitutional AI, RLHF, or safety training methodologies
- Embedding models or vector database implementation details
- Computer use, and vision or image analysis
- Streaming API implementation or server sent events
- Rate limiting, quotas, or API pricing calculations
- Prompt caching implementation details, beyond knowing the feature exists
- Token counting algorithms or tokenization specifics
- Specific cloud provider configurations
Exam details verified against the official exam guide, version 1.0, effective July 2026, and the Anthropic Academy certification page, on September 4, 2026. This is an independent study resource and is not affiliated with, endorsed by, or sponsored by Anthropic.