Building on the Claude API.

You will learn the Claude API from first call to first agent: messages, system prompts, streaming, tool use, and structured output. End state is a small working application — an API integration with at least one tool Claude calls — running against your own key. After this module you can choose the right model (Fable, Opus, Sonnet, Haiku) for a task, estimate and control costs, and build the request patterns every production integration rests on. This is the API leg of Anthropic's Claude Certified Developer exam.

1. Make and debug Messages API calls: roles, system prompts, multi-turn conversations
2. Stream responses and handle stop reasons correctly
3. Define tools and handle the tool-use loop
4. Get reliable structured output for downstream code
5. Pick models deliberately and reason about token costs

• Anthropic, "Using the Messages API" — the request/response shape, stop reasons, and multi-turn structure this module builds on: https://platform.claude.com/docs/en/build-with-claude/working-with-messages
• Anthropic, "Tool use overview" — how tool calls are defined, returned and handled: https://platform.claude.com/docs/en/agents-and-tools/tool-use/overview
• Anthropic, "Choosing a model" and "Pricing" — the two pages to check before committing an architecture to a model: https://platform.claude.com/docs/en/about-claude/models/choosing-a-model and https://platform.claude.com/docs/en/about-claude/pricing
• Anthropic, "Effort" — the primary control over token cost and latency: https://platform.claude.com/docs/en/build-with-claude/effort
• Anthropic courses (API fundamentals, tool use): https://github.com/anthropics/courses
• Anthropic quickstarts: https://github.com/anthropics/claude-quickstarts