// Chuleta rápida · Claude Code

Claude Code Commandsfor serious coding vibes

The actual Anthropic CLI commands you need to stop improvising and start building software with criteria.

40+ commands 6 categories Updated 2026

If you've been doing it for a while vibe coding with Claude Code, you surely already know the command claude and little else. Most people starting out stay there: write a prompt, cross their fingers, and repeat. But Claude Code has over 40 commands and shortcuts designed so you stop improvising and truly control what you build: context, memory, git, sub-agents, permissions, and automations. Save this page and return to it whenever you forget how to undo a change or connect an MCP server.

The 40+ commands, organized by what you really need

01

Getting started and configuration

01claudeStart Claude Code in the current folder
02claude "task"Start with an initial prompt
03/initGenerate the project's CLAUDE.md
04/modelSwitch model (Sonnet, Opus, Haiku)
05/login /logoutChange or log out of account
06/configView and change settings
07/doctorCheck installation health
02

Context and memory

08/memoryEdit CLAUDE.md files from memory
09/add-dirAdd another working directory
10/compactSummarize the conversation (with optional focus)
11/clearClear session history
12/rewindUndo to a previous checkpoint
13/costCheck token consumption
14/exportExport the entire conversation
03

Git and review

15/reviewRequest a code review of the change
16/pr_commentsRead comments on a Pull Request
17"create a commit"Draft and create the commit for you
18"open a PR"Prepare and publish the Pull Request
19/statusAccount and system status
20/bugReport a bug directly to Anthropic
04

Agents, MCP, and tools

21/agentsCreate and manage your own sub-agents
22/mcpConnect MCP servers (Slack, Notion...)
23/hooksAutomate actions on each tool call
24/permissionsWhat Claude can do without asking
25/ideConnect to your editor (VS Code, JetBrains)
26/contextContext window usage breakdown
05

Customization

27.claude/commands/Slash commands tailored for you
28CLAUDE.mdRules Claude always follows
29settings.jsonProject config: permissions and hooks
30/vimActivate vim mode to edit prompts
31/terminal-setupEnable multiline shortcuts in your terminal
32/output-styleChange tone and format of responses
06

Help and control

33/helpGeneral CLI usage help
34Shift + TabChange mode: auto, plan, approval
35Esc EscEdit a previous message in the conversation
36Ctrl+C / Ctrl+DInterrupt task or log out
37--resumeResume a previous session from the terminal
38--continueContinue the last conversation on the fly

★ Bonus: habits of a true vibe coder

·
Plan Mode — think and validate the plan before touching code
·
Sub-agents — distributes independent tasks in parallel
·
"think harder" — ask for more reasoning on difficult tasks
·
A CLAUDE.md — by folder, with module rules
·
Worktrees — multiple branches at once without stepping on each other's toes
·
Small steps — validate each increment before proceeding

The key idea: commands don't make you a better programmer, judgment does

Memorizing this list won't make you a good vibe coder, just like memorizing keyboard shortcuts won't make you a good designer. What makes the difference is when you use each command: stop in plan mode before Claude touches a line of code, split a large problem into several sub-agents instead of chasing a single giant prompt, or leave a CLAUDE.md well-written so the team's criteria survive the conversation.

That layer of judgment — knowing what to ask, when to trust, and when to stop and review — is exactly what we work on thoroughly in the Vibe Coder Course: from idea to a real working product, with Claude Code as a copilot and not a black box.

FROGAMES FORMACIÓN

Ready to move from copying commands to building with judgment?

Vibe Coder Course — from idea to working AI product

Enter the Vibe Coder course

TYPE / TO START · LEARN FROM THE BEST · FROGAMES FORMACIÓN