IntermediateИнженерияclaude

Claude API from A to Z

A complete course on the Anthropic Claude API for developers: Messages API, model selection, prompt engineering for Claude, tool use, streaming, batch API, prompt caching, vision, and production patterns. Real code in TypeScript and Python.

4modules

16lessons

362 mintotal time

Developers integrating Claude into a productaudience

Module 1

Getting Started

Your first API call, the Messages API structure, and choosing a model for the task and budget.

Your first API call

Install @anthropic-ai/sdk, create a client, send a message, and parse the response. Real TypeScript code.

20 min

The Messages API structure

system/user/assistant roles, content blocks (text, image), and how a multi-turn conversation works.

22 min

Choosing a model

Haiku (fast/cheap), Sonnet (balanced), Opus (powerful). When to use which, cost math, model IDs.

20 min

Module 2

Prompt Engineering for Claude

System prompts, the 200K context window, XML tags, and few-shot — what actually works with Claude.

Designing system prompts

Persona, constraints, output format, tone. What works, what doesn't. 3 real examples.

22 min

The 200K context window

What 200K means in practice, how to structure large documents, when chunking beats stuffing everything in.

22 min

XML tags in prompts

<document>, <example>, <instructions>, <thinking>. Why Claude responds better to structured input.

20 min

Few-shot prompting

Module 3

Tool Use

Function calling: defining tools, parsing tool_use, patterns, multi-turn chains, and computer use.

Tool Use basics

Defining tools with JSON Schema, tool_choice, parsing tool_use blocks, sending tool_result back.

25 min

Tool patterns

Web search, database query, code execution, calculator. The full request/response cycle.

25 min

Multi-turn conversations

Building history, when to summarize, token management across turns.

24 min

Computer Use (beta)

What it is, the action types (screenshot, click, type), limitations and latency, when it makes sense.

Module 4

Production

Streaming, the Batch API, prompt caching, vision, and resilient production patterns.

Streaming

stream: true, MessageStreamEvent types, handling text deltas in TypeScript, progressive rendering.

22 min

Message Batches API

When to use it (bulk classification, evals, async jobs), 50% cost saving, how to submit and poll.

24 min

Prompt Caching

cache_control breakpoints, ephemeral, when cache hits happen, cost math. A real example with a large system prompt.

25 min

Vision