Claude Code Commands: Complete Reference and Cheat Sheet (2026)
Every Claude Code slash command, keyboard shortcut, and CLI flag in one place. Includes a printable cheat sheet and how to build custom commands.
Claude Code has more built-in commands than most people use. This is the full list, grouped by what you're trying to do, plus how to add your own.
Quick Answer: Type
/in Claude Code to see available slash commands. The ones worth memorising are/clear(reset context),/compact(summarise and continue),/plan(read-only planning),/mcp(check server connections), and/resume(return to a previous session). Custom commands go in.claude/commands/as markdown files.
Session and context commands
Context management is where most Claude Code sessions go wrong. These are the commands that fix it.
/clear Wipe conversation history, keep the session open
/compact Summarise the conversation and continue with less context
/resume Reopen a previous session
/cost Show token usage and cost for this session
/status Show current model, mode, and working directory
Use /clear when switching to an unrelated task. Use /compact when you're deep in something and running out of context but need the history. The difference matters: /clear loses everything, /compact keeps a summary.
What this fixes: the slow degradation where Claude gets worse over a long session because early context crowds out what's relevant now.
Recommended skills
skill-router-2
by Shippers · 5
Automatically detect, load, and stack the perfect skills combo for any user requ…
AI Feature Eval Writer
by PubsProToolkit
Design and write the eval suite for your LLM-powered feature — the metrics that…
Hooks & Settings for Claude Code
by Markus Isaksson · 19
Master Claude Code's settings hierarchy and hook framework to automate workflows…
Permission and mode commands
/plan Enter Plan Mode for the next prompt (read-only)
Shift+Tab Cycle permission modes: Default → Auto-Accept → Plan
/permissions View and edit tool permissions for this project
Plan Mode blocks every write tool until you approve a plan. Auto-Accept applies edits without asking. Default asks per action.
Full detail in Claude Code Plan Mode.
What this fixes: approving 40 individual edits during a refactor, or the opposite, Claude rewriting eleven files before you noticed.
Configuration and integration commands
/mcp List connected MCP servers and their status
/config Open configuration settings
/model Switch model mid-session
/init Generate a CLAUDE.md for the current project
/memory Edit the project's CLAUDE.md
/mcp is the first thing to run when a server isn't working. It tells you whether the server connected before you start editing config files.
What this fixes: guessing whether a config change took effect.
Keyboard shortcuts
Shift+Tab Cycle permission modes
Ctrl+C Cancel the current operation
Ctrl+D Exit the session
Ctrl+L Clear the terminal display (not the context)
Esc Interrupt Claude mid-response
Esc Esc Edit your previous message
Up / Down Navigate message history
Esc Esc is the most underused shortcut in Claude Code. It lets you rewrite your last prompt and re-run it, instead of adding a correction on top of a bad instruction.
What this fixes: the pattern where a vague first prompt produces a wrong answer and every follow-up inherits the confusion.
CLI flags
claude # Start an interactive session
claude "fix the failing test" # Start with an initial prompt
claude -c # Continue the most recent session
claude -r # Pick a session to resume
claude -p "summarise this repo" # Print mode: run once, output, exit
claude --permission-mode plan # Start in Plan Mode
claude --model claude-opus-4-5 # Start with a specific model
claude mcp list # List configured MCP servers
Print mode (-p) is the one to know for scripting. It runs a single prompt and exits, so you can pipe Claude into shell scripts, git hooks, and CI.
git diff --staged | claude -p "write a conventional commit message"
What this enables: Claude Code as a Unix tool rather than only an interactive session.
Writing custom commands
Custom slash commands are markdown files in .claude/commands/. The filename becomes the command.
Create .claude/commands/review.md:
Review the staged changes.
For each file:
1. Read the full file, not just the diff
2. Check for logic errors, security issues, and missing error handling
3. Rate each finding Critical, Warning, or Suggestion
4. Show a concrete fix for anything Critical
Skip style nits. Focus on things that will break.
Now /review runs that prompt.
Use $ARGUMENTS to accept input:
Write tests for $ARGUMENTS.
Detect the existing test framework from package.json.
Match the naming and assertion style already used in the project.
Cover the happy path, edge cases, and error cases.
Called as /test src/auth.ts.
What this demonstrates: custom commands are prompt templates, not code. Anything you type repeatedly should be one.
Custom commands vs skills
They solve different problems and the distinction matters.
A custom command is a prompt you invoke deliberately. It lives in one project. You type /review and it runs.
A skill is a capability the agent loads automatically when the description matches what you asked. It works across every SKILL.md-compatible agent, Claude Code, Cursor, OpenClaw, Codex CLI.
Use a command for a project-specific workflow you trigger by hand. Use a skill for a method you want applied consistently without remembering to invoke it.
The most installed skills that replace common custom commands: code-reviewer (764 installs, free), git-commit-writer (232 installs, free), and pr-description-writer (85 installs, free).
Browse all Claude Code skills on Agensi.
The cheat sheet
CONTEXT PERMISSIONS CONFIG
/clear /plan /mcp
/compact Shift+Tab /config
/resume /permissions /model
/cost /init
/status /memory
SHORTCUTS CLI
Esc interrupt claude -c continue
Esc Esc edit last claude -r resume
Ctrl+C cancel claude -p print mode
Ctrl+D exit claude mcp list
For the wider Claude Code feature set, read Everything Claude Code.