Claude Agent SDK Reference

Build autonomous AI agents with the Claude Agent SDK.

3|2|Updated Jan 23, 2026
One-click install
npx skills add https://github.com/AeyeOps/aeo-skill-marketplace --skill claude-agent-sdk-reference
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: Claude Agent SDK Reference
Source: https://github.com/AeyeOps/aeo-skill-marketplace/tree/main/aeo-claude/skills/claude-agent-sdk
Command: npx skills add https://github.com/AeyeOps/aeo-skill-marketplace --skill claude-agent-sdk-reference

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill provides a structured reference for building autonomous AI agents using the Claude Agent SDK, enabling developers to design, test, and deploy complex multi-turn workflows with secure permissions and tool orchestration.

Core Features & Use Cases

  • GTVR workflow support: Implement Gather, Take Action, Verify, and Repeat patterns for reliable agent behavior.
  • Persistent and stateless modes: Use both stateless query() and stateful ClaudeSDKClient for multi-turn dialogues and context retention.
  • Tooling and MCP integration: Create custom tools with @tool, expose them via MCP servers, and compose agents from modular skills.
  • Streaming vs batch: Support real-time streaming of responses or batch-style interactions to fit UI needs.
  • Deployment patterns: Outline authentication, permission modes, and logging for production-grade agents.

Quick Start

Install the CLAUDE Agent SDK package and run the included examples to begin building agents. Examples show how to create a client, run a sample query, and extend tools with MCP servers.

Frequently Asked Questions about Claude Agent SDK Reference

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build autonomous AI agents with the Claude Agent SDK?▼

To build autonomous AI agents with the Claude Agent SDK, use Python to implement multi-turn workflows, custom tools, and streaming responses. It requires installing the claude-agent-sdk package and structuring context with a YAML frontmatter in SKILL.md.

What is the GTVR pattern for agent workflows?▼

The GTVR pattern for agent workflows stands for Gather, Take Action, Verify, and Repeat. Implementing GTVR with the Claude Agent SDK ensures reliable behavior in autonomous multi-turn agents by structuring task execution and validation loops.

How do I retain context in multi-turn agent dialogues?▼

You retain context in multi-turn dialogues by using the stateful ClaudeSDKClient API instead of the stateless query() function. This stateful client manages persistent state, enabling continuous multi-turn interactions with the Claude Agent SDK.

Can I integrate custom tools with Claude agents via MCP?▼

Yes, you can integrate custom tools with Claude agents via MCP servers. The Claude Agent SDK allows you to define functions using the @tool decorator and expose them through Model Context Protocol servers for modular agent composition.

Does the Claude Agent SDK support real-time streaming for Python applications?▼

Yes, the Claude Agent SDK supports real-time streaming for Python applications. It provides streaming APIs to deliver responses incrementally, alongside batch-style interactions, allowing you to fit the output method to your UI needs.

What are the limitations of stateless queries in agent workflows?▼

Stateless queries using the query() function do not retain conversation context between calls, limiting their use in continuous multi-turn workflows. For persistent context and stateful multi-turn agent behavior, you must use the ClaudeSDKClient instead.