Skip to Content
View all events

Getting Started with Claude Agent SDK

Published by O'Reilly Media, Inc.

Intermediate content levelIntermediate

Building production-ready agents with Anthropic’s native framework

What you’ll learn and how you can apply it

  • Set up and deploy AI agents using Claude's native framework, primarily in Python, with TypeScript pointers along the way—then run them safely in production with patterns like sandboxing and session resumption
  • Recognize how Claude agents think, reason, call tools, and respond through the complete agent execution cycle
  • Apply task tracking, session persistence, subagents, hooks, and Agent Skills to create sophisticated, context-aware agents
  • Master the SDK's powerful built-in tools, then extend agents with Agent Skills, MCP servers, and custom tools—knowing when to use each

Course description

Create custom agents with Claude Agent SDK, Anthropic's production-ready framework built on the same agent harness that powers Claude Code. The SDK comes with powerful built-in capabilities such as filesystem access, web search, computer use, automatic context management with compaction, and session persistence, allowing you to focus on your agent’s logic rather than its infrastructure. You’ll explore seamless extensibility through Agent Skills and Model Context Protocol (MCP) servers, learning when to leverage built-in capabilities versus building domain-specific extensions.

AI engineer Sajal Sharma takes you through the complete agent loop, helping you create agents with built-in tools, skills, and custom tools; track multistep work with structured task lists; preserve context across runs with session persistence; and delegate to specialized subagents. You'll learn to extend capabilities through MCP servers and Skills, and to deploy agents safely with sandboxing. In four hours, you’ll be equipped to build AI agents that leverage Claude SDK’s production-ready infrastructure for immediate productivity while autonomously accomplishing complex domain-specific, multistep tasks.

This live event is for you because...

  • You’re an AI or ML engineer who’s looking to build production-ready autonomous agents with a framework specifically designed for Claude’s capabilities.
  • You’re a software engineer transitioning to AI development, and you want to understand how to build intelligent tool-using agents without the complexity of general-purpose frameworks.
  • You’re an application developer who wants to integrate Claude’s agentic capabilities into your products with minimal setup and maximum control.
  • You’re an AI researcher or data scientist exploring agentic systems and want hands-on experience with a modern, opinionated agent framework.

Prerequisites

  • Python 3.12+ with pip/uv available to install packages
  • An active Anthropic API key with access to Claude models
  • A code editor (VS Code recommended) and terminal access
  • Intermediate knowledge of programming using Python (TypeScript helpful but not required)
  • Basic understanding of LLMs and AI agents
  • Familiarity with concepts like function calling, tools, and API-based workflows
  • Exposure to agent frameworks (helpful but not required)

Recommended preparation:

Recommended follow-up:

Schedule

The time frames are only estimates and may vary according to how the class is progressing.

Introduction to Claude Agent SDK (75 minutes)

  • Presentation: Core agentic concepts; framework landscape for building agents; introduction to Claude Agent SDK; the SDK's subprocess model; production-ready out-of-the-box features; when to choose Claude Agent SDK; single versus streaming response modes; Python versus TypeScript build options
  • Hands-on exercises: Create a simple agent with minimal configuration in Python; explore single query versus streaming modes
  • Group discussion: What challenges have you faced building agents with various frameworks?
  • Q&A
  • Break

Deep dive into SDK features (90 minutes)

  • Presentation: The Claude agent loop explained; observing and debugging agent reasoning; inspecting message types and tool calls; context management and automatic compaction; session persistence and resumption; query versus continuous conversation modes; built-in tools and their advantages; creating and using Agent Skills; comparison with implementing similar functionality in other frameworks; advanced features
  • Demonstration: The complete agent loop with visible reasoning and tool calling steps
  • Hands-on exercises: Create and use skills with custom workflows and helper scripts
  • Q&A
  • Break

Extending capabilities and multi-agent architecture (75 minutes)

  • Presentation: When and why to extend beyond built-in tools; choosing between Agent Skills, MCP servers, and custom tools; building and using skills and custom tools; extending through Model Context Protocol (MCP); in-process versus external MCP servers; designing effective multi-agent systems; subagent architecture patterns; context isolation strategies; production deployment and sandboxing
  • Hands-on exercises: Integrate custom tools; Build a skills-driven multi-agent system with specialized subagents
  • Q&A

Your Instructor

  • Sajal Sharma

    Sajal Sharma is an AI engineer and technology leader with over eight years of experience in AI/ML, specializing in natural language processing. He works at Menyala, a venture studio in Singapore, where he focuses on building AI-first products and shaping technology strategies. Previously, he led AI initiatives at various consulting and product companies, developing innovative AI solutions across industries. His on-demand course, Building AI Agents with LangGraph, is available on the O’Reilly learning platform. Sajal has delivered a guest lecture at Yale University and has been a mentor for Udacity and the University of Melbourne, where he guided students in machine learning and AI. He holds a master’s degree in information technology from the University of Melbourne.

Skill covered

AI Agents