Claude Certified Architect practice examGet ready for the Claude Architect exam.
Lessons and scored labs on agents, tools, Claude Code, prompts and context, from people who ship with Claude every day.
The lesson plan
36 lessons in 6 chapters · about 5 minutes each
- 27%
1Agents that loop and delegate8
Agentic Architecture & Orchestration
- The Claude tool-use loop Lesson 1 · free
- Coordinators and subagents
- Give subagents the full context
- Run subagents in parallel
- Hooks enforce rules prompts can’t
- Handoffs a human can act on
- Chain prompts or plan adaptively
- Resume, fork, or start fresh
- 18%
2Tools Claude picks correctly6
Tool Design & MCP Integration
- Tool descriptions that route well
- Errors an agent can act on
- Fewer tools per agent
- tool_choice: auto, any, forced
- MCP servers: project or user scope
- Grep, Glob, Read or Edit?
- 20%
3Claude Code for a team6
Claude Code Configuration & Workflows
- CLAUDE.md: user, project, directory
- Skills and slash commands
- Path rules with glob patterns
- Plan mode or direct execution?
- Iterate with examples and tests
- Claude Code in CI pipelines
- 20%
4Prompts that return clean data6
Prompt Engineering & Structured Output
- Explicit criteria beat vague rules
- Few-shot examples that teach
- JSON schemas through tool use
- Validate, retry, feed back errors
- When to use the Batches API
- A second pass catches more
- 15%
5Context and reliability6
Context Management & Reliability
- Keep key facts in long chats
- When to escalate to a human
- Pass errors up, with context
- Explore big codebases, stay focused
- Confidence scores and human review
- Keep every claim’s source
6Scenarios and test day4
Exam Scenarios
- Six scenarios, four on exam
- Four Claude certifications, compared
- Who can take the exam
- Test day
Built on Claude Certified Architect – Foundations (CCAR-F) Exam Guide · 1.0, effective July 2026
Read lesson 1: The tool-use loop
You ask Claude to send an email. Claude can describe that action in text, but sending it takes a tool.
An agent is a loop. You send a prompt, Claude decides what to do, and it can call your tools.
Your prompt joins the system prompt and the tool definitions in the context window.
For a client tool, Claude returns stop_reason "tool_use" and a tool_use block with an id, the tool's name, and its input. Your code runs the tool.
Claude returned a tool_use block. What does your code do next?
- Show it to the user as the answer
- Run the tool and send back a tool_result
- Start a new conversation
Answer: Run the tool and send back a tool_result. Run it, then send a tool_result with the matching id. Claude reads it and decides the next step.
The result goes back. Claude reads it, then calls another tool or answers.
Every turn adds to the context, so keep tool results short and specific.
Next lesson: Coordinators and subagents
The full class, $249 at launch.
Paid once, when it opens. Lesson 1 is free now.
- Scored hands-on labs
- Mock exams
- Lessons on every domain
Only people at Claude Partner Network organizations can sit the exam. Independent prep. Not affiliated with or endorsed by Anthropic.
Questions
Is this from Anthropic?
No. It’s independent prep, not affiliated with or endorsed by Anthropic.
What does it cover?
All five exam domains: agents and orchestration, tools and MCP, Claude Code, prompts and structured output, and context and reliability, in hands-on labs scored on real code, then mock exams.
Who can take the exam?
Per Anthropic Partner Academy, only people at Claude Partner Network organizations can sit it. Check that your organization is a partner before you plan on it.
When does it open, and what will it cost?
Lesson 1 is free now. The full class opens at $249, paid once.