claude-api Skill: Claude API and SDK Reference for Agents

/plugin install claude-api@anthropic-agent-skills

The claude-api skill is a reference for the Claude API and Anthropic SDKs: model IDs, parameters, streaming, tool use, MCP, agents, prompt caching, token counting and model migration. Its job is to stop Claude from answering API questions from memory, which goes stale as models and features change.

What it does

• Loads language-specific references for C#, cURL, Go, Java, PHP, Python, Ruby and TypeScript, plus shared material.
• Triggers when a prompt mentions Claude or Anthropic models or SDKs, or when the task is LLM-shaped (agents, tool definitions, RAG, summarization) and no other provider is in use.
• Steps aside when another provider’s SDK is present in the project.

Install

In Claude Code you already have it: /claude-api is a bundled skill, and it activates automatically when your code imports anthropic or @anthropic-ai/sdk. Subcommands start specific workflows, for example:

/claude-api migrate
/claude-api cost-optimize
/claude-api build-eval

For other environments, or to pin the open repo version, install it from the anthropics/skills marketplace:

/plugin marketplace add anthropics/skills
/plugin install claude-api@anthropic-agent-skills

See the skills guide for bundled versus installed skills.

Example usage

Add prompt caching to our summarization endpoint and explain the expected savings
Migrate this service to the newest Claude model and update deprecated parameters
Build a tool-use loop in TypeScript that calls our search function

Tips

• Let the skill check model IDs and parameters instead of pasting old snippets.
• Use cost-optimize before scaling an integration, and build-eval before changing prompts in production.
• For agentic coding work in Claude Code itself, see getting started.

Source and license

Source: anthropics/skills/skills/claude-api. License terms are in the skill’s LICENSE.txt.