claude-api

Preserve prompt caching during model migrations by applying centralized guidance for switching Claude surfaces and model versions while maintaining existing cache logic. This ensures backend services retain cache hits across Anthropic SDK updates.

Choosing between Claude API surfaces involves routing logic based on task complexity, using single calls for direct prompts, tool use for structured actions, and agentic loops for multi-step reasoning. Managed Agents handle complex workflows automatically.

Implement prompt caching and streaming with the Anthropic SDK by applying best-practice defaults and workflow patterns for backend integrations. This Skill provides robust caching, streaming, and error handling logic for Claude API applications.

Yes, you can use Managed Agents for backend services requiring multi-surface routing. They provide automated handling of complex agent workflows, switching between single-call Claude API, tools, and agent loops based on application requirements.

The Anthropic SDK supports error handling and tool use within agentic workflows through official SDK features. This Skill ensures proper usage of these capabilities alongside prompt caching and streaming for robust backend integrations.

Avoid prompt caching in Claude API integrations when working with highly dynamic, non-repetitive prompts that offer minimal cache hit opportunities. Caching provides the most value for stable system prompts and recurring context in backend services.