Claude API — Anthropic API Usage Guide#

Purpose#

Provide guidance on using the Anthropic API, selecting Claude models, leveraging tool use, and following best practices.

Topics#

Model Selection#

• Claude Opus 4: Most capable model. Best for complex reasoning, coding, and multi-step tasks.
• Claude Sonnet 4: Balanced performance and speed. Good default for most tasks.
• Claude Haiku 3.5: Fastest and most cost-effective. Best for simple tasks and high-volume use.

API Basics#

• Base URL: https://api.anthropic.com/v1/messages
• Authentication: x-api-key header with your API key
• Required headers: anthropic-version: 2023-06-01, content-type: application/json
• Key parameters: model, messages, max_tokens, system

Tool Use (Function Calling)#

• Define tools in the tools array with name, description, and input_schema
• The model returns tool_use content blocks when it wants to call a tool
• Send tool results back as tool_result content blocks
• Support for parallel tool calls and sequential chains

Best Practices#

• Use the system prompt for persistent instructions and context
• Set max_tokens appropriately to avoid truncation
• Use streaming (stream: true) for better user experience on long responses
• Implement exponential backoff for rate limit errors (429)
• Cache system prompts with prompt caching to reduce costs
• Use stop_sequences to control output format

Common Patterns#

• Multi-turn conversation: Maintain the full message history
• Structured output: Use tool use to get JSON responses
• Extended thinking: Enable thinking for complex reasoning tasks
• Batch API: Use /v1/messages/batches for high-volume async processing

Guidelines#

• Always check the latest API documentation for current model names and features.
• Include practical code examples in Python or JavaScript when explaining concepts.
• Mention pricing considerations when recommending models.