Claude Code

Claude Code is Anthropic’s CLI-first AI coding agent. Unlike IDE-embedded tools, it runs in your terminal, works with any editor, scripts into CI/CD pipelines, and gives you full control over every interaction through hooks, subagents, and headless mode. If you spend your day in the terminal, Claude Code meets you where you already work.

You probably already know it can edit files and run commands. Power users pull 3x more out of it through config, context engineering, and workflow tricks the official docs never surface — patterns that came out of Anthropic’s own engineering, data science, and non-technical teams, out of developers who migrated from other AI coding tools, and out of production work on large codebases. This guide collects them, from first-run setup to enterprise-scale monorepo strategy.

Why CLI-first matters

Claude Code is not a chatbot bolted onto an editor. It is an agentic system that reads your codebase, runs shell commands, edits files, and chains complex multi-step operations — all from a single terminal session. This architecture unlocks patterns that IDE-only tools cannot replicate:

Scriptable automation: Pipe output, chain commands, run headless in CI
Hook-driven guardrails: Block dangerous operations, auto-format code, inject context — deterministically, not probabilistically
Subagent delegation: Spawn isolated agents for parallel tasks, each with their own context window
Session persistence: Resume conversations across days, share context through CLAUDE.md files
Zero lock-in: Works with VS Code, Cursor, Neovim, JetBrains, or no editor at all
Adaptive reasoning: Toggle it with Option+T/Alt+T and tune it with /effort; fixed token caps are a 4.6 compatibility path

Where each part of the Claude Code guide lives

Advanced Techniques

Deep dives into hooks, custom commands, memory management, enterprise proxies, LLM gateways, GitHub Actions, and cost monitoring. For developers who want programmatic control over every aspect of Claude Code.

Start with advanced techniques

Productivity Patterns

Battle-tested workflows for terminal productivity, prompt engineering, batch operations, script automation, IDE coordination, debugging, code review, and deployment. Ship faster with less friction.

Explore productivity patterns

Everything Else

The rest of the guide follows the shape of the work: quick start, planning, testing, shipping, operating in production, team rollout, and the reference material behind all of it.

The 100 tips, by category

Start here: What Claude Code Can Do That You Haven't Tried A curated companion to the 100 tips — ~40 features power users miss, from Gui Ferreira's NDC AI 2026 talk. Read this first, then work through the categories below.

Tips 1-15: Setup and Configuration Essential setup steps, IDE integration, permission configuration, and initial project structure

Tips 16-25: CLAUDE.md Optimization Master the art of creating effective CLAUDE.md files for persistent memory and project context

Tips 26-35: Command Line Mastery Essential CLI commands, slash commands, keyboard shortcuts, and terminal productivity

Tips 36-50: Large Codebase Management Strategies for navigating, understanding, and modifying enterprise-scale codebases

Tips 51-65: Workflow Optimization Transform your development process with AI-first workflows and productivity patterns

Tips 66-75: Performance and Cost Management Optimize token usage, manage costs effectively, and maximize performance

Tips 76-85: Advanced Techniques Master extended thinking modes, MCP integration, custom hooks, and parallel workflows

Tips 86-95: Team Collaboration Best practices for team adoption, shared configurations, and collaborative development

Tips 96-100: Troubleshooting and Best Practices Common issues, security considerations, and proven patterns for success

The habits that separate power users

Five themes run through all 100 tips, and they matter more than any individual setting:

Treat Claude Code as a development partner. The most successful developers think of it as a “very fast intern with perfect memory” — eager to help, incredibly capable, but needing clear direction and verification.
Context is everything. Output quality correlates directly with the context you provide through CLAUDE.md files, clear prompts, and structured workflows.
Embrace the terminal interface. Initially skeptical users consistently report that the terminal becomes their primary development environment, not a secondary tool.
Automate everything possible. From permission management to testing workflows, the more you automate, the more time you have for creative problem-solving.
Keep improving the setup itself. Have Claude Code update its own documentation and encode your patterns, so the environment gets better every week.

The ten tips with the biggest payoff

Use /clear frequently - Start fresh for each new task to save tokens (Tip #29)
Create hierarchical CLAUDE.md files - Build persistent memory at multiple levels (Tip #17)
Use message queuing - Type multiple prompts and let Claude work through them (Tip #53)
Leverage extended thinking - Toggle it with Option+T/Alt+T, tune adaptive models with /effort, and use ultrathink as a one-turn deeper-reasoning hint; positive MAX_THINKING_TOKENS values apply only to 4.6 fixed-budget compatibility mode (Tip #76)
Run multiple instances in parallel - Work on different parts of your codebase simultaneously (Tip #41)
Install the GitHub CLI - Wire up gh for PR creation and review from the terminal (Tip #4)
Create custom slash commands - Automate repetitive workflows (Tip #35)
Monitor costs with /cost - Track usage and optimize for efficiency (Tip #66)
Use --dangerously-skip-permissions only in sandboxed or CI environments - Skip prompts where a wrong command can’t hurt you, never on your main checkout (Tip #3)
Reserve Auto mode for well-defined, test-guarded tasks - acceptEdits only skips edit/common-filesystem prompts; Auto uses a safety classifier for broader unattended work and still needs review (Tip #51)

Quick reference

Need	Go to
Set up hooks and automation	Hooks System Mastery
Write better prompts	Prompt Engineering
Control costs	Monitoring and Costs
Use Claude Code in CI/CD	GitHub Actions
Manage large codebases	Large Codebase Tips
Configure CLAUDE.md	CLAUDE.md Mastery

Where to start with Claude Code

If you are new, work through the categories in order, starting with Setup and Configuration — each config tweak and workflow pattern stacks on the last, so the payoff compounds. If you are already productive, jump straight to the problem in front of you: Advanced Techniques for hooks and parallel workflows, Workflow Optimization for the daily loop. If you are rolling Claude Code out to a team, share the sections that match each person’s role rather than the whole list at once.