Claude Code (CLI)

What it is

Claude Code is Anthropic’s agentic coding tool. The CLI is the terminal version: you cd into a repo, run claude, and Claude reads files, edits them, runs tests, and commits — on demand and with your approval. It’s the same engine that powers Claude Code Desktop and the web app at claude.ai/code; the CLI is where the team ships first and where every knob is exposed.

Where to find it

Install with one of the following, then run claude:

Your setup	Command
macOS / Linux / WSL	`curl -fsSL https://claude.ai/install.sh \| bash`
Windows (PowerShell)	`irm https://claude.ai/install.ps1 \| iex`
Homebrew	`brew install --cask claude-code`
npm (Node 18+)	`npm install -g @anthropic-ai/claude-code`

Don’t use sudo with the npm install — it causes permission issues later. Verify with claude --version.

How to use it

First-day flow

cd path/to/your/project and run claude.
On first launch, /login authenticates via browser (Enterprise/SSO) or use ANTHROPIC_API_KEY / Bedrock / Vertex env vars.
Ask a question. Claude reads files on demand — nothing to upload.
When Claude proposes an edit you’ll see a diff. Choose Yes, Yes, don’t ask again, or No.
Run /init once per project to generate a starter CLAUDE.md capturing conventions and commands.

Permission modes

Press Shift+Tab to cycle: default → acceptEdits → plan (and auto if you launched with --enable-auto-mode). Plan mode is read-only; use it to let Claude outline big changes before touching files.

CLAUDE.md — project memory

Drop a CLAUDE.md at your repo root. Claude reads it automatically at the start of every session. Keep it under ~200 lines. Include build/test/lint commands, naming conventions, hard constraints, and gotchas. Leave out full API docs, changelogs, and aspirational rules. Prompt caching keeps the cost low.

Slash commands worth knowing

/clear — wipe conversation, keep project memory. Start fresh between unrelated tasks.
/compact — summarize conversation to free context mid-task.
/model — switch between Sonnet (default), Opus (deeper reasoning), Haiku (fastest/cheapest). Use /model opusplan to plan with Opus and execute with Sonnet.
/cost — token and dollar usage for this session (API billing).
/usage — Enterprise plan limits and rate-limit status.
/permissions — pre-approve safe commands per project.
/agents, /hooks, /skills — customize subagents, lifecycle hooks, and packaged workflows.
/doctor — diagnostic report for support tickets.

Full keybindings: ? inside a session.

Good to know

Models: Sonnet for most coding, Opus when stuck or for wide changes, Haiku for quick mechanical edits. /model is always the source of truth for what’s available to your account.
Usage meter depends on auth. Enterprise seats get a pool that resets on a rolling window. API-key auth (Console, Bedrock, Vertex, Foundry) is pay-per-token with no hard stop.
1M context is available in Claude Code for Max, Team, and Enterprise users on Opus 4.6 or 4.7.
Worktrees: claude --worktree launches an isolated git worktree session. Run 3–5 in parallel. /batch distributes large migrations across many worktree agents.
Chrome extension & Desktop app close the verification loop for web code — Claude can read console logs, errors, and DOM state. Power-user tip #1 is to give Claude a way to verify its own work.
Teleport and remote control (Pro/Max/Team/Enterprise, CLI v2.1.51+) move sessions between terminal, desktop, web, and mobile.
Troubleshooting quick hits: command not found → reopen terminal. EACCES on npm install → use the native installer or fix npm prefix. /login hangs → paste the URL manually. 403 on every request → /model to a model your seat has access to. Last resort: claude doctor.

What’s changed recently

2026-04-18 — Page created from official help-center articles.