Anthropic
Claude Code
1 hour, 56 minutes
CC
Course Description
Learn Claude Code for free! Customize Claude Code for your codebase, using CLAUDE.md, plan mode, and permissions that adhere to your team’s standards. Build reusable skills tailored to your processes and wire up hooks so Claude behaves consistently across the whole team. Go under the hood with Claude Code to generate and ship higher-quality code.
Prerequisite: Knowledge of
Prompt Engineering and basic
software development experience.
Preview
Course Details
Published: May 20, 2026
Learn Straight from the Experts Who Shape the Modern Web
Your Path to Senior Developer and Beyond
- 250+ In-depth courses
- 24 Learning Paths
- Industry Leading Experts
- Live Interactive Workshops
Table of Contents
Introduction
Section Duration: 16 minutes
Lydia Hallie from Anthropic’s Claude Code team introduces Claude Code, an AI toolset designed for developers. Lydia shares her extensive development journey and transition into AI. This course will focus on Claude Code's fundamentals, components, and usage.
Lydia explains how Claude Code works under the hood, focusing on its architecture, model options, and interaction loop. She highlights the distinction between the Claude Code harness and the AI models, the agentic loop process, and the trade-offs between different models for software engineering tasks.
Basics
Section Duration: 1 hour, 11 minutes
Lydia emphasizes the importance of a Claude.md file for improving AI model understanding of a codebase and streamlining coding workflows. The `/init` skill will generate a new Claude.md file for a project and provide context to AI models, improving their ability to generate relevant code and reduce unnecessary tool calls.
Lydia demonstrates how to manage Claude Code permissions. She covers the Plan Mode, Allow List, Ask Mode, and Deny Mode permission controls and shares best practices for each mode.
Lydia introduces the concept of "effort" and highlights how different effort levels affect the behavior of AI models. Higher effort increases inference costs but can improve output quality. Lydial also explores the advisor tool, which allows users to designate one model for planning and another for implementation.
Lydia works with "skills" in Claude Code to automate repetitive, multi-step workflows. Skills are Markdown files that define specific procedures and can be customized for each project. They can even inject dynamic context by running shell commands before sending content to the model.
Lydia uses the Skill Creator to generate a new skill. This helps automate skill setup and testing by running evaluations comparing code performance with and without the skill, generating HTML reports to assess effectiveness. Lydia also highlights some third-party skill repositories and marketplaces.
Lydia introduces hooks in Cloud Code, explaining how they enable custom logic execution at specific points in an agentic lifecycle. Hooks provide better control over tool execution, can trigger commands, HTTP endpoints, or prompts, enhancing automation and safety. Hooks are configured in settings.json.
Lydia explains how sub-agents run in separate loops with their own context, tools, and system prompts. They operate independently from the main conversation, preventing context pollution. Multiple sub-agents can run in parallel, handling tasks like code review or exploration simultaneously. Results from sub-agents are returned to the main agent without introducing intermediate clutter.
Lydia highlights the difference between sub-agents and agent teams in AI workflows. The agent teams enable inter-agent communication and collaboration, unlike independent sub-agents. They are ideal for complex, multi-step workflows, whereas sub-agents thrive in environments with simpler tasks.
Other Features
Section Duration: 28 minutes
Lydia introduces plugins that bundle skills, hooks, agents, and other components into reusable packages. Plugins have a specific file structure, including a .claude-plugin file and a hooks.json for hooks.
Lydia briefly explains the Model Context Protocol (MCP), a standardized interface that enables AI tools like Claude to interact seamlessly with external systems beyond code repositories. She highlights how MCP simplifies AI integration by providing a uniform protocol rather than relying on diverse APIs, making it easier to integrate with tools such as issue trackers.
Lydia demonstrates a Claude Desktop workflow with GitHub. Claude Code Desktop offers an intuitive interface with session logs, preview servers, and no need for browser extensions. Users can edit UI components by sending screenshots and component data to the AI model. Diff views and PR creation are integrated, enabling easy review and collaboration. Auto mode and permission settings streamline AI-assisted code edits.
Lydia briefly introduces Cowork, a productivity tool that automates daily tasks, communications, and schedules through integration with various OS-level applications. It can automate routine tasks such as reviewing missed messages, emails, and calendar events at the end of each day, and allows setting specific prompts and routines that run at scheduled times.
Lydia discusses the benefits of the Agent SDK for interacting with the Claude Code runtime. It requires an API key for access and provides the same runtime capabilities, including querying and programmatic tool creation.
Wrapping Up
Section Duration:
Lydia wraps up the course, summarizing the flexibility and adaptability of Claude Code and encourages engineers to reach out with feedback on how the Claude Code team can improve the tool.