201+ commands, flags, shortcuts and settings across 18 buckets. Search it, slide the density, filter by bucket, and click any command to copy it.
Install it and launch in your project.
curl -fsSL https://claude.ai/install.sh | bash
macOS / Linux — recommended
brew install --cask claude-code
Homebrew on macOS
irm https://claude.ai/install.ps1 | iex
Windows PowerShell
npm install -g @anthropic-ai/claude-code
npm (requires Node.js 18+)
cd your-project && claude
Launch Claude Code
claude update
Update to the latest version
Pick the brain and how hard it thinks.
/model fable
claude-fable-5 — flagship for the hardest, longest tasks
/model opus
claude-opus-4-8 — complex reasoning & architecture
/model sonnet
claude-sonnet-4-6 — daily coding
/model haiku
claude-haiku-4-5 — fast & efficient for simple tasks
/model best
Auto-picks the strongest available model
/model opusplan
Opus in plan mode, Sonnet for execution
opus[1m]
Append [1m] for a 1M-token context window
/effort low|medium|high|xhigh|max
Set reasoning effort
ultrathink
Drop anywhere in a prompt for one-off deep reasoning
/fast on (⌥O)
Faster Opus output — not a smaller model
Drive the session without leaving home row.
Bash mode (run a shell command)
Mention a file or folder
Run a slash command
Insert a line break
Interrupt Claude mid-response
Open the rewind / edit-previous menu
Clear input (press again to exit)
Exit session
Clear & redraw the screen
Toggle transcript / thinking view
Reverse-search command history
Toggle the background task list
Background a running bash command
Paste an image from the clipboard
Open the prompt in $EDITOR
Kill all background agents (twice to confirm)
Cycle permission modes
Switch model
Toggle extended thinking
Toggle fast mode
Quick-launch Claude in supported IDEs
Interactive session control, one slash away.
/clear
Start a fresh conversation (keeps memory)
/compact [focus]
Summarize the conversation to free context
/resume [session]
Resume by ID or name
/branch [name]
Fork the conversation here (alias /fork)
/context [all]
Visualize context-window usage
/export [file]
Export the conversation as text
/copy [N]
Copy the last (or Nth) response to clipboard
/recap
One-line session summary
/btw <question>
Side question without bloating history
/model [model]
Switch model (← → adjusts effort)
/effort [level]
/fast [on|off]
/plan [desc]
Enter plan mode
/goal [condition]
Keep working until a condition is met
/code-review [effort] [--fix]
Review the diff for bugs + cleanups; --fix applies
/simplify [target]
Cleanup-only review (no bug hunting)
/security-review
Scan the diff for vulnerabilities
/review [PR]
Review a pull request locally
/run
Launch & drive the app to confirm a change works
/verify
Confirm a code change behaves as intended
/diff
Interactive diff viewer
/loop [interval] [prompt]
Run a prompt on a recurring interval
/workflows
View & manage background multi-agent runs
/batch <instruction>
Decompose a change across parallel worktrees
/schedule [desc]
Create a scheduled cloud agent (cron routine)
/config
Open settings (alias /settings)
/permissions
View or update tool permissions
/mcp
Manage MCP servers & OAuth
/agents
Create & manage subagents
/hooks
Configure lifecycle hooks
/skills
List available skills
/reload-skills
Re-scan skills without restarting
/plugin
Manage plugins (install / enable / disable)
/status
Version, model, account & connectivity
/usage
Plan limits + per-category cost (alias /cost)
/doctor
Diagnose installation & configuration
/help
Show available commands
Everything you can pass to `claude`.
claude "query"
Start with an initial prompt
claude -p / --print
Headless mode: run via SDK and exit
claude -c / --continue
Continue the most recent conversation
claude -r / --resume "<session>"
Resume a session by ID or name
claude --model <alias>
Set model (fable/opus/sonnet/haiku/id)
claude --effort <level>
claude --fallback-model <m>
Auto-fallback when primary is unavailable
claude --safe-mode
Disable customizations for troubleshooting
claude --permission-mode <m>
default/acceptEdits/plan/auto/dontAsk/bypassPermissions
claude --dangerously-skip-permissions
Skip all permission prompts ⚠️
claude --add-dir <path>
Grant extra working-directory access
claude --bg --exec '<cmd>'
Run a shell command as a background agent
claude -w / --worktree <name>
Start in an isolated git worktree
claude --mcp-config <file>
Load MCP servers from JSON
claude doctor
Settings files and the config CLI.
claude config list
List the current configuration
claude config get <key>
Check a specific setting
claude config set <key> <value>
Change a setting value
claude config add <key> <value>
Push to an array setting
claude config remove <key> <value>
Remove from an array setting
/etc/claude-code/managed-settings.json
Enterprise — highest priority
.claude/settings.local.json
Project local (personal, gitignored)
.claude/settings.json
Project shared (team settings)
~/.claude/settings.json
User global (personal defaults)
Tune behaviour from the shell.
ANTHROPIC_API_KEY
API key for direct authentication
ANTHROPIC_MODEL
Default model (overridden by --model)
ANTHROPIC_BASE_URL
Override the API endpoint (proxy / gateway)
CLAUDE_CODE_EFFORT_LEVEL
Effort: low/medium/high/xhigh/max/auto
MAX_THINKING_TOKENS
Max extended-thinking tokens (0 = off)
ENABLE_PROMPT_CACHING_1H
Opt into 1-hour prompt-cache TTL
CLAUDE_CODE_SAFE_MODE
Enable safe mode
CLAUDE_CODE_DISABLE_1M_CONTEXT
Turn off the 1M-token context window
CLAUDE_CODE_DISABLE_CRON
Disable scheduled tasks
CLAUDE_CODE_USE_BEDROCK
Route through AWS Bedrock
CLAUDE_CODE_USE_VERTEX
Route through Google Vertex AI
BASH_DEFAULT_TIMEOUT_MS
Default bash timeout (default 120000)
MCP_TIMEOUT
MCP server startup timeout (ms)
DISABLE_TELEMETRY
Disable telemetry collection
CLAUDE.md, rules, and auto-memory.
/init
Generate a starter CLAUDE.md from the codebase
/memory
Edit CLAUDE.md, toggle auto-memory, view entries
@import
Pull files into CLAUDE.md: See @README (max 4 hops)
./CLAUDE.md
Project memory (team)
./CLAUDE.local.md
Project memory (personal, gitignored)
~/.claude/CLAUDE.md
User memory (all projects)
.claude/rules/*.md
Path-scoped rules via a paths: frontmatter glob
~/.claude/projects/<proj>/memory/
Claude's persistent auto-memory notes
Keep under ~200 lines
Longer CLAUDE.md reduces context & adherence
An automatic safety net for edits.
Esc Esc
Open the rewind menu
/rewind
Rewind code and/or conversation
/undo
Alias for /rewind
/checkpoint
Conversation rewind
Restore to a previous message, keep the code
Code rewind
Revert file changes, keep the conversation
Full rewind
Restore both code and conversation
Not tracked: bash / external
rm/mv/cp & manual edits can't be undone
Roll your own reusable commands.
.claude/commands/
Project commands (team shared)
~/.claude/commands/
Personal commands (all projects)
$ARGUMENTS
Placeholder for all arguments
$1 $2 $3
Positional arguments
!`command`
Run a command before processing
@filename
Include file contents
argument-hint:
Frontmatter hint text
model:
Frontmatter model override
allowed-tools:
Restrict tool access
/mcp__server__prompt
Invoke an MCP prompt
Programmatic, scriptable Claude.
claude -p "query"
Non-interactive — print the result and exit
claude -p --output-format json
Output JSON with metadata
claude -p --output-format stream-json
Stream JSON as it arrives
claude -c -p "query"
claude --resume <id> -p
Resume a specific session
echo "query" | claude -p
Process piped input
claude --verbose -p
Enable verbose logging
claude --max-turns 3 -p
Limit agentic turns
Model-invoked modular capabilities.
.claude/skills/
Project skills
~/.claude/skills/
Personal skills
SKILL.md
Skill definition + optional scripts
name:
Unique skill identifier
description:
When / why Claude should use it
Extend Claude Code from marketplaces.
Open plugin management
/plugin marketplace add <url>
Add a plugin marketplace
/plugin install <name>@<mkt>
Install from a marketplace
/plugin enable <name>@<mkt>
Enable a disabled plugin
/plugin disable <name>@<mkt>
Disable without uninstalling
/plugin uninstall <name>@<mkt>
Completely remove a plugin
Wire in external tools & data.
claude mcp add <name> <command>
Add an MCP with a command
claude mcp add --transport sse <name> <url>
Add an SSE transport
claude mcp add --transport http <name> <url>
Add an HTTP transport
claude mcp list
List installed servers
claude mcp remove <name>
Remove a server
mcp__github
Approve all tools from a server
mcp__github__get_issue
Approve a specific tool (no wildcards)
Parallel, isolated development.
git worktree add <path> <branch>
Create with an existing branch
git worktree add <path> -b <branch> <start>
Create with a new branch
git worktree list
List all worktrees and branches
git worktree remove <path>
Remove a worktree directory
git worktree prune
Clean up worktree references
Start Claude in an isolated worktree
Specialized assistants with their own context.
.claude/agents/
Project subagents (team shared)
~/.claude/agents/
User subagents (all projects)
Open the subagent management interface
claude "Use the [name] subagent to..."
Explicitly invoke a subagent
Subagent identifier
When the subagent should be invoked
tools:
Allowed tools (inherits all if omitted)
inherit / opus / sonnet / haiku
Decide what runs without asking.
default
Ask for sensitive operations
acceptEdits
Auto-accept edits, ask for dangerous ops
plan
Propose changes, you approve before execution
auto
Auto-approve low-risk operations
bypassPermissions
Skip all checks ⚠️
Bash(npm run *)
Allow specific bash patterns
Read(./.env*)
Control file read access
Bash(rm -rf:*)
Deny dangerous commands
allow / ask / deny
Permission list actions
Run scripts on lifecycle events.
PreToolUse
Before tool calls (can block)
PostToolUse
After tool calls complete
UserPromptSubmit
When you submit a prompt
Stop
When Claude finishes responding
SubagentStop
When subagent tasks complete
PreCompact
Before a compact operation
SessionStart
When a session starts / resumes
SessionEnd
When a session ends
Open hooks configuration
Exit 0 / Exit 2
0 = success, 2 = block & show to Claude