Claude BlueprintCCAR-F prep, unofficial
Start chapter 1.1

For the Claude Certified Architect exam

AI agents are invisible.
Claude Blueprint draws them.

An animated study book for the Claude Certified Architect exam. Watch every request and tool call happen, step by step.

Most exam guides describe this. We show it.

Scroll, and watch one tool call make the round trip the whole exam is built on.

  1. Claude sits behind a wall.

    It reads text and writes text. It can't open a file, call an API or check the weather.

  2. Every request carries everything.

    Your code packs the whole conversation and slides it in. The API remembers nothing between calls.

  3. Claude asks. It never acts.

    To get the weather it writes a tool_use block: a tool name, an input, and an id.

  4. Your code does the work.

    It calls the real API, then sends back a tool_result with the same id so Claude can match them.

  5. It loops until Claude is done.

    stop_reason: end_turn. One small field decides what your code does next, and the exam leans on it hard.

Don't just watch. Predict.

Every board stops after each step. Tap Next when you're ready, and at the moments that matter it asks you what happens next before it shows you.

Try it here. This is the real thing, straight from chapter 1.1.

The exam doesn't ask what a tool_use block is. It asks what your system should do next.

Built for judgment
Every board ends on a decision, and every exam-style question explains why each wrong option is wrong.
Honest about change
Claude moves fast. Where today's docs differ from what the exam expects, the chapter says so.
Built in the open
An unofficial concept, not affiliated with Anthropic. Original questions only, no leaked material.

Every idea in Domain 1 is a board you can step through. Open one.

The whole exam, by weight.

Five domains. Domain 1 is the biggest, and it's ready now. The rest are being written.

27%D1 · Agentic Architecture & Orchestration8 chapters ready
18%D2 · Tool Design & MCP Integrationbeing written
20%D3 · Claude Code Configuration & Workflowsbeing written
20%D4 · Prompt Engineering & Structured Outputbeing written
15%D5 · Context Management & Reliabilitybeing written
  1. 1.1The Agentic Loop
  2. 1.2Workflows vs. Agents
  3. 1.3Coordinator–Subagent Orchestration
  4. 1.4Spawning Subagents and Passing Context
  5. 1.5Enforcement Gates and Structured Handoff
  6. 1.6Hooks in Depth
  7. 1.7Task Decomposition Strategies
  8. 1.8Sessions: Resume, Fork, and Stale Context

Start with the agentic loop.

Twenty minutes, ten pages, and you'll never picture a tool call the same way.

Start chapter 1.1

Unofficial study material. Not affiliated with, or endorsed by, Anthropic. Claude is a trademark of Anthropic. Registering for the real exam needs a Claude Partner Network company email.