Lesson Content

The Anthropic API – what it is.

The Anthropic API provides programmatic access to Claude models. Your application sends a request (typically a JSON payload with a model ID, system prompt, and user messages) and receives a response (text or structured output). The API is stateless – each request includes all the context needed to generate the response.

Core components.

Models: Anthropic offers multiple Claude models across capability tiers. As of 2026, the main families include Opus (highest capability), Sonnet (balanced), and Haiku (fastest/lowest cost). Current model IDs and capabilities are at docs.anthropic.com/en/docs/about-claude/models – always use current model IDs from the documentation rather than ones memorized from a course.

Messages API: The primary API endpoint. Accepts a structured messages array (system + user/assistant turns), model ID, and parameters. Returns a completion.

Tokens: The unit of API pricing and context length. Tokens are approximately 0.75 words in English. Both input (prompt) and output (response) tokens are counted and billed. Understanding token consumption is essential for cost management.

Tools: The mechanism for function calling – allowing Claude to request your application to run a function and return results. Covered in depth in a later course.

SDKs: Anthropic provides official Python and TypeScript/JavaScript SDKs. The SDKs wrap the HTTP API with type-safe methods, automatic retry logic, and other production conveniences. Using an SDK is recommended for most integrations.

Documentation resources.

Primary developer reference: docs.anthropic.com

• API Reference: complete endpoint documentation
• Model overview: current model IDs, capabilities, context windows, pricing
• Guides: prompt engineering, tool use, production best practices
• Anthropic Cookbook (GitHub): practical code examples