Prepare the certification Claude Certified Developer — Foundations

The Messages API

The core request/response contract: request anatomy (model, max_tokens, system, messages), statelessness and sending full history each turn, multi-turn conversations, assistant prefills, stop_reason semantics, and streaming responses.

5 lessons · 300 questions

Tool Use and Function Calling

Giving Claude tools: defining tools with JSON schemas, the tool_choice modes (auto / any / forced), the request→tool_use→execute→tool_result loop, parallel tool calls in one turn, and returning results and errors (the isError flag).

5 lessons · 300 questions

Structured Output and Prompt Techniques

Getting reliable machine-readable output: JSON via tool_use schemas, required vs optional/nullable fields to prevent fabrication, enums with escape hatches, few-shot examples, prefills, and validation with retry-on-feedback.

5 lessons · 300 questions

Performance, Cost, and Reliability

Building efficient, resilient integrations: prompt caching for repeated context, the Message Batches API for high-volume async work, token counting and model selection, and handling rate limits with retries and backoff.

5 lessons · 300 questions

Integrating MCP and the SDKs

Wiring Claude into real systems: the official Anthropic SDKs, connecting MCP servers, tools vs resources vs prompts, .mcp.json configuration with environment-variable secrets, and testing/evaluating LLM-backed features.

5 lessons · 300 questions