Claude Certified Architect Foundations

The concept. Claude wrote a book about Claude, out of the documentation Anthropic publishes about itself. The recursion is the fun part. The useful part is that the platform documents sit on several sites, filed by feature, with nothing joining them, and the exam asks why the architecture works the way it does. Answering that means holding all of it in one head at once.

How it was built. The certification site, the platform documentation and the official exam guide came down into local vaults, then went back together as one argument. Every claim traces to the page it came from. Claude Code changes constantly, so check anything time-sensitive before you bet an exam question on it.

What’s inside. The five exam domains, arranged the way the exam actually thinks. The agentic loop as control flow. Hub-and-spoke as the default multi-agent shape. What makes the model pick one tool over another. How CLAUDE.md files merge, and where hooks let you interrupt. Prompting as constraint specification rather than persuasion. Context discipline, and what an error does crossing an agent boundary. Each chapter names the point where the decision forks.

It will not teach you to use Claude Code. It assumes you have, and works on the part the exam tests: why any of it is built that way.