Guides

A real first-day setup for Claude Code that goes beyond install. What to set up, what to skip, how to think about the first session, and the things most people miss in week one.

Which terminal to use on macOS, the Option-key problem nobody warns you about, how to wire up notifications, and the other small settings that make Claude Code feel like it was built for your machine rather than just running on it.

Native Windows is genuinely supported now and you don't need WSL the way you used to. The honest guide to picking PowerShell or WSL2, what Git for Windows changes, the install pitfalls that catch people, and the small terminal settings that matter.

The handful of workflow features that actually change how much you get out of Claude Code, why most users discover them too late, and the patterns Anthropic's own teams use to keep sessions productive across long tasks.

CLAUDE.md is the single highest-leverage file in a Claude Code project, and most people write it badly. The honest guide to what belongs in it, where it lives, how the new auto memory system works alongside it, and the mistakes that quietly hurt response quality.

The configuration surface that turns Claude Code from a useful default into a tool genuinely shaped to your workflow. Settings files, the permission rule system, when to use hooks vs CLAUDE.md vs skills, and the patterns that survive long-term.

The two main ways to extend Claude Code are skills (custom slash commands you write in markdown) and MCP servers (external services that expose tools to the model). Most users discover these too late. Here's the honest guide to when to reach for each.

Everything you do in the interactive terminal, Claude Code can also do unattended. Three ways to run it in a pipeline: the headless CLI (claude -p), the Agent SDK, and the official GitHub Action. Same engine, three entry points, one set of cost and permission rules to get right.