Exam guide: Claude Certified Architect – Foundations
Who the Claude Certified Architect – Foundations exam is for
According to its exam guide, this certification validates that practitioners can make informed decisions about tradeoffs when implementing real-world solutions with Claude. It tests foundational knowledge across four technologies: Claude Code, the Claude Agent SDK, the Claude API and the Model Context Protocol (MCP). The ideal candidate described in the guide is a solution architect who designs and builds production applications with Claude, typically with six months or more of practical experience with these tools.
Access is restricted. Anthropic's "Who can take an exam" page states that Claude certification exams are open to people who work at organisations in the Claude Partner Network, and that registration requires a company email on a domain listed in the organisation's partner record. Candidates must be 18 or older and hold a government-issued ID. Outside the partner network, the public Claude Academy courses end with a quiz and a certificate of completion, which is not a certification.
Exam format
| Parameter | Value |
|---|---|
| Exam code | CCAR-F (exam guide version 1.0, effective July 2026) |
| Number of items | 60 |
| Item format | Multiple choice and multiple response; each item states how many responses to select |
| Structure | 4 scenarios drawn at random from a bank of 6 |
| Time limit | 120 minutes |
| Delivery | Pearson VUE, online proctoring or test center |
| Passing score | Scaled score of 720 on a scale of 100 to 1,000 |
| Results | Pass or fail with the scaled score, plus the percentage of correct answers by domain |
| Validity | 12 months from the date the credential is awarded |
The score is criterion-referenced: you are measured against a fixed standard, not against other candidates. The percentages by domain help you see weak areas but do not decide the result.
The six scenarios
Every question sits inside a production scenario. The exam presents four of the six scenarios listed in the guide:
- Customer support resolution agent: an Agent SDK agent with backend tools exposed through MCP, which must know when to escalate.
- Code generation with Claude Code: team workflows, CLAUDE.md, custom commands, plan mode.
- Multi-agent research system: a coordinator delegating to specialised subagents; the system produces cited reports.
- Developer productivity: an agent that explores codebases with the built-in tools and MCP servers.
- Claude Code in continuous integration: automated reviews, test generation, pull request feedback.
- Structured data extraction: extraction from unstructured documents, validated against JSON schemas.
Since you do not know which four will come up, prepare all six.
Domains and weights
| Domain | Weight | Main topics |
|---|---|---|
| Agentic Architecture & Orchestration | 27 % | Agentic loops and stop reasons, coordinator and subagents, context passing, multi-step workflows, Agent SDK hooks, task decomposition, session resumption and forking |
| Tool Design & MCP Integration | 18 % | Tool descriptions, structured errors from MCP tools, distributing tools across agents, MCP servers in Claude Code, built-in tools |
| Claude Code Configuration & Workflows | 20 % | CLAUDE.md hierarchy, commands and skills, path-specific rules, plan mode, iterative refinement, CI/CD |
| Prompt Engineering & Structured Output | 20 % | Explicit criteria, few-shot examples, structured output with tool use and JSON schemas, validation and retry, batch processing, multi-pass review |
| Context Management & Reliability | 15 % | Long conversations, escalation, error propagation, large codebases, human review and confidence, information provenance |
The first domain carries more weight than any other, and it is a primary domain in three of the six scenarios.
Exam guide and current documentation
The exam guide is version 1.0 from July 2026, and the product documentation keeps moving. For example, the guide describes custom slash commands in .claude/commands/, while the Claude Code documentation now states that custom commands have been merged into skills. The guide calls the subagent tool "Task"; the Agent SDK documentation now calls it the Agent tool, while it still appears as "Task" in the system:init tools list. Learn the guide's terms, then check the current documentation for each topic you study.
How to prepare
- Read the exam guide in full and turn each task statement into a checklist item.
- Check your eligibility on the "Who can take an exam" page before planning anything else.
- Take the free prep courses listed by Anthropic, such as Building with the Claude API and Introduction to Model Context Protocol.
- Build a small agent with the Agent SDK: a complete loop that checks the stop reason, calls tools, handles errors and hands work to a subagent.
- Configure Claude Code on a real project: a CLAUDE.md hierarchy, rules scoped to paths, one skill and one MCP server.
- Build an extraction pipeline that forces a JSON schema through tool use, validates the result and retries on errors, then try the Message Batches API on it.
- Practise with our Claude Certified Architect practice exam under timed conditions, then review each missed question in the free study mode.
Practice Exam: Claude Certified Architect – Foundations
25 free questions with explanations, no account needed.