Claude Code CLI Basics

slash commands

Commands starting with / control the session itself, rather than being a request for Claude to do something in your code.

Command	Does
/help	lists available commands and basic usage
/clear	starts a fresh session, discarding the current conversation's context
/compact	summarizes the conversation so far to free up context — use before you hit the limit, not after
/config	opens configuration settings
/init	generates a starter CLAUDE.md by having Claude read and summarize the codebase

Projects and users can also define their own custom slash commands — a skill (see Claude Skills) is one way these get packaged and shared.

running shell commands directly

Prefixing input with ! runs it as a literal shell command in the current session rather than a message to Claude — handy for something you want to run yourself and have the output land directly in the conversation (e.g. an interactive login command Claude shouldn't run on your behalf).

permission rules

Claude Code asks before running a tool call that isn't already allowed. Rules take the shape Tool(pattern) and live in an allow/deny list — deny always wins over allow, and a more specific pattern overrides a general one.


{
  "permissions": {
    "allow": [
      "Bash(npm run test *)",
      "Read(~/.zshrc)"
    ],
    "deny": [
      "Bash(curl *)",
      "Read(./.env)",
      "Read(./secrets/**)"
    ]
  }
}

where configuration lives

Location	Scope	Shared?
`~/.claude/settings.json`	you, every project	no
`.claude/settings.json`	this project	yes — commit it
`.claude/settings.local.json`	you, this project only	no — gitignore it
Managed policy settings	every user in the organization	yes, IT-deployed

When the same rule appears at multiple levels, managed policy always wins, then command-line flags, then local, then project, then user settings — except permission allow/deny rules specifically, which merge across every scope instead of the higher one overriding the lower. Run /status in a session to see exactly which settings files actually loaded.

a few settings worth knowing about

Key	Does
model	sets the session's model — see Choosing a Claude Model
autoMemoryEnabled	toggles auto memory (on by default) — see Effective Prompting
hooks	event-triggered scripts — see Hooks & Automation
env	environment variables set for every tool call in the session

reference

code.claude.com — overview
code.claude.com — settings