CCA Practice
Exam
Claude Certified Architect — Foundations
How This Works
This is independent practice material, not an official Anthropic exam or a validated predictor of certification results.
This practice exam has 20 scenario-based questions across all 5 domains, with 15 minutes on the clock. The local practice target is 720/1000 (15 of 20 correct). The score is a simple percentage multiplied by 1000, not official exam scaling.
These fictional practice scenarios illustrate production engineering decisions. The distractors are drawn from the anti-patterns seen in agent workflows. Answers describe these scenarios, not universal rules for every architecture. Consult current vendor documentation before implementation.
Reference: Agent SDK subagents, tool definitions, and batch processing, and Lost in the Middle. Reviewed September 27, 2026.
// Domains tested
// Sample question — try before you commit
// Scenario
In this fictional scenario, your customer support agent uses the Messages API and a client-side refund tool. During testing, you observe that the agent occasionally returns a text response to the user while also including a tool_use block requesting the process_refund tool in the same API response.
What should your agentic loop code do when it receives this response?