Language:JapaneseEnglish

claude-api Claude Code Skill

Author
Anthropic
Topics
AI Agents & AI App Development · Web Search, Research & Learning · Migration & Modernization
License
See LICENSE.txt
First cataloged
2026-07-09 (UTC)
Explanation last updated
2026-08-21 (UTC)
Source (GitHub) last updated
2026-08-21 (UTC) (6 days ago)

The explanation below is AI-generated. Please verify it against the sources.

This is a Claude skill called claude-api that acts as a reference guide for writing code against the Claude API and the official Anthropic SDKs. According to the SKILL.md, it activates whenever a prompt mentions Claude or Anthropic, or when a task involves an LLM without a stated provider, and it steps aside when another provider such as OpenAI or Gemini is already in use. It documents current model identifiers, pricing, request parameters, streaming, tool use, Model Context Protocol integration, agent-building approaches, prompt caching, token counting, and steps for migrating code to newer Claude models. The skill also defines subcommands for migrating existing code to a new model, auditing prompts for outdated patterns, and upgrading the Python SDK across a major version. It is organized as per-language reference files (Python, TypeScript, Java, Go, Ruby, C#, PHP, and raw HTTP/cURL) that Claude is instructed to read before generating SDK code.

About the service

The reference material describes the Anthropic Claude API, a hosted service reachable through a Messages endpoint that provides chat-style model access, tool calling, batching, file uploads, token counting, and a separate Managed Agents surface for server-hosted, stateful agents. It also documents access to Claude through third-party cloud platforms including Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.

What you can do with claude-api

  • Look up current Claude model identifiers, context window sizes, and per-token pricing
  • Get guidance on choosing between a single API call, a tool-use workflow, or Managed Agents for a given task
  • Receive SDK code guidance for Python, TypeScript, Java, Go, Ruby, C#, PHP, or raw HTTP requests
  • Run built-in subcommands such as migrate (move code to a newer model), prompt-audit (flag outdated prompt patterns), and upgrade (move the Python SDK across a major version)
  • Get reference details on streaming, prompt caching, extended thinking/effort settings, task budgets, token counting, and structured outputs
  • Get guidance on authentication methods, provider-specific clients (Bedrock, Vertex AI, Foundry), and common implementation pitfalls such as error handling and tool-result formatting

Sources

History of claude-api

  • Claude Code Skill Updated skill content (SKILL.md) of claude-api
  • Claude Code Skill Updated skill content (SKILL.md) of claude-api
  • Claude Code Skill Updated skill content (SKILL.md) of claude-api
  • Claude Code Skill Updated skill content (SKILL.md) of claude-api
  • Claude Code Skill Updated skill content (SKILL.md) of claude-api

Back to list