Jump to learning, practice, or reference tools.
Search the full Claude Code feature list when you need to find where a command, hook, tool, or setting lives.
Need the short printable version instead? Open the cheat sheet .
230 of 230 features
Show help information and learn available commands
Ask a side question without adding it to the conversation context
Configure Chrome integration for browser automation
Clear conversation history to start fresh or reduce context
Open interactive diff viewer to review changes
View or edit Claude Code configuration settings
Show current session status and state
List available subagents and delegation options
List available skills with type-to-filter search and auto-invoke triggers
List configured hooks for debugging event-driven automation
Analyze session patterns for optimization
Install the Claude Slack app for team communication integration
Customize keyboard shortcuts including chord key support
List connected MCP servers and check external integrations
View loaded memory files to debug context loading
Generate a mobile QR code for accessing Claude Code on mobile
View usage passes and subscription information
Manage plugins: install, remove, update, and list extensions
Enter planning mode for complex implementations using read-only tools
Rewind to a previous checkpoint to undo changes or explore alternatives
Save and restore conversation checkpoints
Show token usage costs to monitor spending
Show context window usage to manage conversation length
Export the current conversation for external reference
Configure extra usage limits for rate limit management
Submit feedback or bug reports to Anthropic
Authenticate with Anthropic to access features
Sign out of Claude Code to switch accounts
Toggle sandbox mode for safe command execution in restricted environments
Toggle vim mode for vim-style text editing in the prompt
Run diagnostics to troubleshoot Claude Code issues
Reload installed plugins without restarting Claude Code
Show release notes to check new features
Enable remote control for accessing Claude Code sessions via API
Manage permissions to control tool access and authorization
Manage sessions for multi-session workflows
Rename the current session to organize work
Resume a previous session to continue work
View and manage the todo list to track tasks
View and manage background tasks and async operations
Copy the last response to clipboard for quick sharing
Transfer the current session to another machine to continue work remotely
Open or switch to the Claude Desktop app
Change the color theme to customize appearance
Unified dashboard merging /cost and /stats — shows cost estimate, duration, and code change metrics
Fork the current conversation to explore alternatives without losing history
Show session statistics to review metrics
Configure the status line display
View session stickers earned as fun rewards
Toggle fast output mode to speed up responses
Configure terminal integration features including iTerm2 clipboard access
Check for Claude Code updates and manage version upgrades
Analyze code for performance optimization opportunities
Prepare a pull request with context and descriptions
Generate API documentation from code
Create a git commit with contextual message generation
Stage, commit, and push changes in one quick step
Restructure documentation to improve organization and clarity
Setup CI/CD pipeline configuration for new projects
Expand test coverage by generating additional unit tests
Standard interactive mode that prompts for each tool call authorization
Auto-accept file edits while prompting for other tool types
Read-only tool access for planning and exploration without writes
Fully autonomous operation that accepts all tools without prompting (Research Preview)
Skip all permission checks for CI/CD and headless environments
Skip tools that would require permission for non-interactive scripting
Multi-step research and complex multi-file task agent with all tools available
Implementation planning agent for architecture design using read-only tools
Codebase exploration agent using Haiku 4.5 for quick searches and code understanding
Command execution agent for git operations and terminal tasks
Status line configuration agent powered by Sonnet 4.6
Help and documentation agent using Haiku 4.5 for learning Claude Code features
Comprehensive code quality review agent for code review sessions
Feature architecture design agent for planning new features
Deep codebase analysis agent for understanding existing features
Code review agent applying Clean Code principles for maintainability
Test strategy and coverage planning agent for comprehensive test plans
Technical documentation agent for API docs and guides
Security-focused code review agent for security audits
Full feature implementation agent for end-to-end feature development
Root cause analysis agent for bug investigation
SQL queries and data analysis agent for data tasks
Comprehensive code review skill auto-invoked by 'Review this code' or 'Check quality'
Brand consistency checker for marketing copy and content
API documentation generator auto-invoked by 'Generate docs' or 'Document API'
Systematic code refactoring skill auto-invoked by 'Refactor this' or 'Clean up code'
Review code for quality and simplicity after writing code
Run prompts on multiple files for batch operations
Debug failing tests or errors during debugging sessions
Run prompts on a recurring interval for scheduled or repeating tasks
Build applications with Claude API for API development tasks
PR review workflow plugin bundling 3 commands, 3 agents, and GitHub MCP
Deployment and monitoring plugin bundling 4 commands, 3 agents, and Kubernetes MCP
Documentation generation suite bundling 4 commands, 3 agents, and templates
MCP server for PR management, issues, and code access via GitHub API
MCP server for SQL queries and direct database access
MCP server for advanced file operations beyond built-in file tools
MCP server for team communication, notifications, and Slack updates
MCP server for document access, editing, and review via Google Docs
MCP server for project management and task tracking via Asana
MCP server for payment data access and financial analysis
MCP server providing persistent memory for cross-session recall
Built-in MCP server for up-to-date library documentation lookup
Triggered on session begin or resume for setup tasks
Triggered when CLAUDE.md or rules file is loaded for custom instruction handling
Triggered before prompt processing for input validation and transformation
Triggered before any tool execution for validation and logging
Triggered when permission dialog is shown for custom approval flows
Triggered after successful tool execution for formatting and notifications
Triggered after tool execution failure for error handling and logging
Triggered when Claude sends a notification for external alerting
Triggered when a subagent is spawned to initialize subagent context
Triggered when a subagent finishes for chaining subsequent actions
Triggered when Claude finishes responding for cleanup and reporting
Triggered on API error ending a turn for error recovery and logging
Triggered when a teammate agent is idle to distribute work across agent teams
Triggered when a task is marked complete for post-task processing
Triggered when a task is created via TaskCreate for task tracking and logging
Triggered when configuration is updated to react to settings changes
Triggered when the working directory changes for directory-specific setup
Triggered when a watched file is modified for file monitoring and rebuilds
Triggered before context compaction for state preservation
Triggered after context compaction completes for post-compact actions
Triggered when a git worktree is created for worktree environment setup
Triggered when a git worktree is removed for worktree resource cleanup
Triggered when an MCP server requests user input for input validation
Triggered when a user responds to an MCP elicitation for response processing
Triggered on session termination for cleanup and saving state
PreToolUse:Bash hook that validates shell commands before execution
PostToolUse:Write hook that scans written files for security issues
PostToolUse:Write hook that auto-formats code after file writes
UserPromptSubmit hook that validates prompts before processing
Stop hook that tracks token usage across sessions
PreToolUse:Bash hook that runs pre-commit validation checks
PostToolUse:Bash hook that logs all executed shell commands
Organization-managed policies automatically applied to all users for org-wide standards
Project-level memory file for team standards and project context, committed to git
Modular project rules directory for organized, team-shared Claude instructions
User-level memory file for personal preferences and global instructions
Modular personal rules directory for organized user-specific Claude instructions
Local machine-specific memory file that is git-ignored for personal overrides
Automatically captured insights and corrections stored across the session
Control Claude Code sessions remotely via API by sending prompts and receiving responses programmatically
Run Claude Code in a browser-based environment via Anthropic Console
Native desktop application for Claude Code with enhanced UI
Coordinate multiple agents working on related tasks with shared context
Background task management and monitoring for async operations
Context-aware command suggestions that appear automatically based on current context
Isolated git worktrees for safe parallel branch development
Isolated execution environments for running commands safely in restricted containers
Block specific domains from outbound network access even when a broader allowedDomains wildcard would permit them
OAuth authentication for MCP servers for secure credential-based access
Search and discover MCP tools dynamically across connected servers
Set up recurring tasks using /loop and cron tools for automated scheduling
Browser automation with headless Chromium via --chrome flag or /chrome command
Customize keybindings including chord key support via /keybindings or ~/.claude/keybindings.json
Fully autonomous operation without permission prompts via --mode auto (Research Preview)
Multi-channel communication via Telegram, Slack, and other platforms (Research Preview)
Voice input for prompts using microphone icon or voice keybinding
Hook type that spawns a subagent instead of running a shell command by setting type: agent
Hook type that injects prompt text into the conversation by setting type: prompt
MCP servers can request user input during tool execution via Elicitation events
WebSocket-based transport for MCP server connections via transport: websocket config
Language Server Protocol integration via plugins for editor features like autocomplete
Organization-managed drop-in configurations auto-applied to all users via managed policies (v2.1.83)
Override the model for a specific subagent (e.g., haiku-4.5 for faster, cheaper tasks)
Set reasoning effort level (low, medium, high) for a subagent to control compute usage
System prompt injected at subagent start to customize agent behavior
Explicitly deny specific tools from a subagent for security and scope control
YAML frontmatter field defining trigger phrases for automatic skill invocation
YAML frontmatter field setting reasoning effort level for the skill (low, medium, high)
YAML frontmatter field specifying the shell to use for skill scripts (bash, zsh, sh)
Alias for /loop — run a prompt on a recurring interval for proactive monitoring or polling
Create, update, list, or run cloud-backed scheduled routines conversationally
Top-level manifest key pointing at background Monitor tool configs that auto-arm when the plugin is enabled
Tool that creates an isolated git worktree and switches into it; accepts a path parameter to enter an existing worktree instead
Experimental dual-model feature that lets a faster executor model consult a higher-intelligence advisor model mid-task; enable with /advisor
On native macOS/Linux builds, Glob and Grep tools are replaced by embedded bfs and ugrep binaries through Bash for faster searches
Settings.json option controlling how long session files are retained before automatic deletion at startup; also controls orphaned subagent worktree cleanup when the worktree has no uncommitted changes, no untracked files, and no unpushed commits (worktrees created with --worktree are excluded; default: 30 days)
MCP servers connect concurrently at startup by default for faster initialization when multiple servers are configured
Environment variable to enable forked subagents that inherit full conversation context from the main session, changes /fork behavior, and runs all subagent spawns in the background (interactive mode only)
Character-wise (v) and line-wise (V) visual selection in the input editor with operators like d, y, c, p, and text objects
Create and switch named color themes stored as JSON in ~/.claude/themes/; plugins can ship themes too
Hook type that invokes an MCP tool directly — call external services from hooks without shelling out
PostToolUse and PostToolUseFailure hook inputs include duration_ms with tool execution time in milliseconds
Environment variable that blocks all update paths including manual claude update — stricter than DISABLE_AUTOUPDATER
Include "$defaults" in autoMode.allow, soft_deny, or environment to extend built-in rules instead of replacing them
Settings.json option that points the footer PR badge at a custom code-review URL instead of GitHub
Environment variable that hides the working directory from the startup logo banner
The --from-pr flag now accepts GitHub, GitHub Enterprise, GitLab merge-request, and Bitbucket pull-request URLs
Tag a plugin version for release with optional push, dry-run, and force options
Remove unused or orphaned plugin files and dependencies to keep the installation clean
Exempt an MCP server from tool search deferral so its tools are always loaded into context
Environment variables to override default model names for Haiku, Sonnet, and Opus model tiers
Auto-accept confirmation prompts in CLI commands like plugin install and project purge
Run /ultrareview non-interactively from CI or scripts; prints findings to stdout (--json for raw output) and exits 0 on success or 1 on failure
Skills can reference the current effort level with ${CLAUDE_EFFORT} in their content to adapt behavior dynamically
PostToolUse hooks can replace tool output for all tools via hookSpecificOutput.updatedToolOutput, not just MCP tools
Environment variable to select a Bedrock service tier (default, flex, or priority), sent as the X-Amzn-Bedrock-Service-Tier header
Managed setting that disables inline shell execution in skills, custom slash commands, and plugin commands
Environment variable set for subprocesses so tools like gh can attribute traffic to Claude Code
Managed policy setting to restrict plugin marketplaces by hostPattern (domain) and pathPattern (repository path)
Delete all Claude Code state for a project including sessions, memory, and cached data; supports --dry-run, -y/--yes, -i/--interactive, and --all flags
CLI flag equivalent to --permission-mode bypassPermissions that skips all permission prompts; name emphasizes the security trade-off
Authentication workflow for environments where the browser callback cannot reach localhost (WSL2, SSH, containers); manually paste the OAuth code after browser sign-in
When --dangerously-skip-permissions is active, writes to .claude/skills/, .claude/agents/, and .claude/commands/ are still allowed without prompting
Cap the maximum runtime of claude ultrareview in minutes before returning partial results
Settings.json option controlling which git reference new worktrees branch from—'fresh' (default, from remote default branch) or 'head' (from local HEAD)
Managed-only settings (Linux/WSL2) to specify absolute paths to the bubblewrap and socat binaries for sandbox isolation, overriding automatic PATH detection
Admin-tier managed setting controlling whether parent-supplied settings (from SDK or IDE) merge with or are overridden by admin-deployed policies—'first-wins' (default) or 'merge'
Environment variable available in hook commands containing the current effort level, allowing hooks to adapt behavior based on session effort
Environment variable set in Bash subprocesses with the current effort level (low, medium, high, xhigh, max), enabling scripts to adjust behavior dynamically
Environment variable set in Bash subprocesses and hooks containing the unique session identifier for tracking and correlation
Environment variable that disables the fullscreen alternate-screen renderer, keeping conversation in the terminal's native scrollback
Status bar hint that appears while an image paste (Ctrl+V) is being read from the clipboard
CLI flag that fetches a plugin as a .zip archive from a URL for the current session, allowing dynamic plugin loading without permanent installation
Environment variable that force-enables synchronized output rendering on terminals where auto-detection fails
Environment variable enabling automatic background updates for Homebrew and WinGet installations with a restart prompt when a new version is available
Environment variable that enables automatic population of the model picker from a configured LLM gateway's /v1/models endpoint
Managed-only setting that enables channel notifications (Telegram, Slack) for the organization, required for API key authentication on Console
Environment variable containing a GitHub personal access token, used by the gh CLI and cloud session tools for repository access
Environment variable specifying the GitHub hostname for GitHub Enterprise Server installations, used by the gh CLI in cloud sessions
Environment variables for authenticating with GitHub Enterprise Server instances, used by the gh CLI when GH_HOST points to an enterprise server
Set a completion condition that Claude works toward across turns, with a live overlay showing elapsed time, turn count, and token usage
Show a plugin's component inventory (skills, agents, hooks, MCP servers) and projected token cost per session
Open the agent view showing a roster of all Claude Code sessions with state (working, waiting, completed, failed, idle, stopped) and activity
When args is set as an array, the hook command spawns directly without a shell (exec form); when omitted, the command passes to shell for tokenization (shell form)
Environment variable that disables the background agents and agent view UI
Environment variable that caps the number of agentic turns in a session, useful for limiting automated runs
Environment variable that shows the terminal's native block cursor instead of Claude Code's default cursor appearance
