Claude Code Best Practices: A Quick Reference Guide
A curated collection of tips for getting the most out of Claude Code.
## Setup Your CLAUDE.md
Create a CLAUDE.md file at your project root. Include:
- Build and test commands
- Code style guidelines
- Repository conventions and patterns
Treat CLAUDE.md like a prompt—refine it iteratively. Use emphasis keywords like "IMPORTANT" or "YOU MUST" to improve adherence.
## The Explore-Plan-Code-Commit Workflow
1. **Explore**: Have Claude read relevant files first
2. **Plan**: Ask for a detailed plan before coding (use Shift+Tab twice for Plan Mode)
3. **Code**: Let Claude implement with your approval
4. **Commit**: Have Claude create clear commit messages
## Key Tips
**Be specific upfront.** Instead of "add tests," specify which edge cases to cover. Clear instructions reduce course corrections.
**Use tab completion.** Reference specific file paths to guide Claude to the right context.
**Course correct early.** Press Escape to interrupt. Double-tap Escape to edit previous prompts. Don't let Claude go down rabbit holes.
**Clear context between tasks.** Use `/clear` to prevent irrelevant conversation from filling the context window.
**Provide visuals.** Paste screenshots or drag images—Claude excels with visual context.
## Advanced: Run Multiple Instances
Use `git worktree add` to create separate working directories, then run multiple Claude instances in parallel for independent tasks.
## Advanced: MCP Servers
Configure Model Context Protocol servers to extend Claude's reach beyond your terminal—connect to external APIs, databases, or custom tools.
---
*Curated from Anthropic's official documentation and community best practices.*