00
Overview
Claude Code is an AI assistant designed to help you write, understand, and improve code. It works alongside you, learning from your project to provide contextual help across your whole development workflow.
01
What it does
Understands
Reads and explains your codebase, making complex architectures accessible.
Creates
Writes new features from your requirements and existing patterns.
Refines
Identifies and resolves issues while maintaining code quality.
Tests
Validates functionality to ensure reliability.
Adapts
Works autonomously on extended tasks while respecting your preferences.
02
Getting started
Configure your environment. Create a claude.md file in your project root — Claude's guide to your workflow. Include your development practices, coding standards, business-specific terminology, and architectural decisions. Run /init to generate a starter file, then refine it.
Plan Mode
Claude develops a strategy, creates a task list, and waits for your approval.
Shift + TabAuto Mode
Claude works independently, making changes as it proceeds. Best for routine tasks with clear requirements.
Enable Extended Thinking (Tab) to give Claude more reasoning time for complex problems — especially effective combined with Plan Mode for architectural decisions.
03
Essential features
Checkpoints
Your work is preserved before each significant change. If something goes wrong, press Escape twice to return to a previous state.
Subagents
Specialized assistants for focused tasks — build validation, test execution, code review, architecture planning. Each keeps its own context.
/agentsHooks
Automated responses to workflow events — format after edits, notify on completion, validate before execution, maintain standards.
Custom Commands
Shortcuts for frequent tasks. /review instead of a long instruction. Define your own vocabulary for common workflows.
04
Best practices
Discovery
Let Claude explore your codebase to understand structure and patterns — especially valuable on new projects or unfamiliar components.
Planning
Use Plan Mode with Extended Thinking for complex features. Provide design docs; let Claude surface edge cases and improvements.
Implementation
Write tests first. Claude's quality is consistently high when it can validate its work against defined criteria — and variable without it.
Context management
Claude performs best with focused information. Use /compaction, deploy subagents, be selective about file reads, and let it keep its own notes.
05
Common situations
Agrees too readily
Add instructions to claude.md encouraging critical feedback; ask it to challenge assumptions and flag issues.
Incomplete understanding
Spend more time in Plan Mode, add business context, and share design docs for complex features.
Excessive file creation
Redirect when noticed. This behavior keeps improving with each model iteration.
Avoiding hard problems
Use Extended Thinking, provide encouragement and direction, and break the problem into smaller steps.
06
Quick reference
/init
Create the claude.md file.
/compaction
Compress the conversation.
/agents
Configure subagents.
/memory
Add persistent context.
Shift + Tab
Toggle Plan Mode.
Tab
Toggle Extended Thinking.
Escape × 2
Rewind changes.
07
Integration
- Terminal — command-line interface for direct interaction.
- VS Code — native sidebar integration with your editor.
- JetBrains IDEs — terminal-based with editor awareness.
- GitHub Actions — automated code review on pull requests.
- Custom applications — embed via the Claude Agents SDK.
★
Philosophy
Start simple. Build gradually. Write tests. Plan carefully. Review thoroughly.
Claude is your partner in development. You guide the direction; Claude handles the details.