opensource/claude-code-guide
3
0
Fork 0
mirror of https://github.com/zebbern/claude-code-guide.git synced 2026-09-19 05:23:28 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
22abba7fa74e99733bc2208045aa2e1cc429d951
claude-code-guide/README.md
T
zebbern 22abba7fa7 Update README.md
2025-06-23 05:17:19 +02:00

30 KiB
Raw Blame History

Visit: https://zebbern.github.io/ for more detailed info!

Section Status
Guides on how to install on Windows, Linux, MacOS ✅
Tips and Tricks ✅
MCP Overview with what to use ✅
Community Guides ✅
Troubleshooting ✅
How to use Claude code the most optimal way ✅

I Usually Start my Claude with Claude --dangerously-skip-permissions

Claude Code - Complete Command Reference

Claude Code Status License

The most comprehensive Claude Code command reference available 📚

Discover hidden commands, advanced flags, and power-user techniques that most developers don't know about.

Table of Contents

Quick Start

Essential Commands (Click to expand) 
# Interactive Mode
claude                      # Start interactive REPL
claude "your question"      # Start with initial prompt

# One-Shot Mode  
claude -p "analyze this"    # Quick query and exit
cat file | claude -p "fix"  # Process piped content

# Management
claude config              # Configure settings
claude update              # Update to latest
claude mcp                 # Setup MCP servers

Core CLI Commands

Interactive Commands

Command Description Example
claude Start interactive REPL claude
claude "query" REPL with initial prompt claude "help me debug this"
claude -p "query" One-off query (print mode) claude -p "explain this function"
cat file | claude -p Process piped content cat error.log | claude -p "summarize"

Management Commands

Command Description Use Case
claude config Configure settings Initial setup
claude update Update to latest version Stay current
claude mcp Configure MCP servers Add integrations

CLI Flags & Options

Essential Flags

Flag Short Description Example
--print -p Print response without interactive mode claude -p "help"
--resume -r Resume specific session by ID claude -r abc123
--continue -c Load most recent conversation claude -c
--verbose - Enable detailed logging claude --verbose
--debug -d Enable debug mode claude --debug
--help -h Show help information claude --help

Output Format Control

Format Options
Flag Options Description
--output-format text, json, stream-json Control response format
--input-format text, stream-json Control input format
--json - Shorthand for JSON output

Examples:

# Text output (default)
claude -p "help" --output-format text

# Structured JSON
claude -p "analyze" --output-format json

# Streaming JSON for real-time
claude -p "process" --output-format stream-json

Permission & Security Flags

⚠️ Warning: Use security flags with extreme caution!

Security Controls
Flag Risk Level Description
--dangerously-skip-permissions HIGH Skip ALL permission prompts
--allowedTools <tools> LOW Whitelist specific tools
--disallowedTools <tools> LOW Blacklist specific tools

Safe Examples:

# Allow only git operations
claude --allowedTools "Bash(git:*)"

# Allow file editing only
claude --allowedTools "Edit,View"

# DANGEROUS - skip all checks
claude --dangerously-skip-permissions

Advanced/Hidden Flags

Expert-Level Options

System Control

Flag Purpose Example
--mcp-debug Debug MCP connections claude --mcp-debug
--max-turns <n> Limit agentic interactions claude --max-turns 5
--add-dir <path> Add working directories claude --add-dir ../lib ../src
--model <model> Set specific model claude --model sonnet

Prompt Engineering

Flag Purpose Example
--system-prompt Override system prompt claude --system-prompt "You are an expert"
--append-system-prompt Append to system prompt claude --append-system-prompt "Focus on security"
-system Alternative syntax claude -system "Custom prompt"

Tool Management

Flag Purpose Example
--permission-prompt-tool Custom permission handler claude --permission-prompt-tool auth_tool
--allow-tool Allow specific tool claude --allow-tool Edit
--disallow-tool Block specific tool claude --disallow-tool Bash

Experimental Flags

Under Development: These flags are proposed/experimental

Future Features
Flag Status Purpose
--no-memory Proposed Disable CLAUDE.md loading
--no-tools Proposed Exclude tool context
--no-env Proposed Disable environment context

Cloud Provider Integration

Provider Flag Service
Amazon Bedrock --use-bedrock AWS
Google Vertex AI --use-vertex GCP 
# Use AWS Bedrock
claude --use-bedrock "analyze this code"

# Use Google Vertex AI
claude --use-vertex "help with deployment"

Configuration Commands

Config Management

Command Purpose Scope Example
claude config list List all settings Current claude config list
claude config get <key> Get specific value Current claude config get model
claude config set <key> <value> Set value Project claude config set model sonnet
claude config set -g <key> <value> Set global value Global claude config set -g model opus
claude config add <key> <value> Add to list Current claude config add allowedTools Edit
claude config remove <key> <value> Remove from list Current claude config remove allowedTools Bash

Quick Setup:

# Set global preferences
claude config set -g model claude-sonnet-4
claude config set -g verbose true

# Set project-specific settings
claude config set allowedTools "Edit,View,Bash(git:*)"

MCP Commands

MCP: Model Context Protocol for extending Claude's capabilities

MCP Server Management
Command Purpose Example
claude mcp serve Start MCP server claude mcp serve
claude mcp add <name> <cmd> Add stdio server claude mcp add git "git-mcp-server"
claude mcp remove <name> Remove server claude mcp remove git

Popular MCP Servers:

# Database integration
claude mcp add postgres "postgres-mcp-server"

# Web browsing
claude mcp add browser "browser-mcp-server"

# File system operations
claude mcp add fs "filesystem-mcp-server"

Slash Commands

Interactive Mode Only: These commands work inside the Claude REPL

Core Slash Commands

Essential Commands
Command Purpose Quick Tip
/help Show all commands Start here!
/clear Clear conversation Fresh start
/status System information Check everything
/cost Token usage stats Monitor spending
/exit Exit safely Clean shutdown
Management Commands
Command Purpose When to Use
/config View/modify settings Setup & tuning
/permissions Manage tool access Security control
/doctor Health check Troubleshooting
/init Create CLAUDE.md New projects
/bug Report issues Found a problem?
History & Session Commands
Command Purpose Pro Tip
/undo Revert last change Instant rollback
/compact Compress conversation Save context
/login Switch accounts Multi-account
/logout Sign out Security
Interface Commands
Command Purpose For Users Who
/vim Enable Vim keybindings Love Vim

Advanced Slash Commands

Workspace Management
Command Purpose Example
/add-dir <path> Add working directory /add-dir ../backend
/memory Edit memory files /memory
/model <name> Change AI model /model claude-opus-4
Integration Commands
Command Integration Purpose
/ide IDE Tools Manage IDE integrations
/mcp MCP Servers Manage MCP connections
/install-github-app GitHub Setup GitHub Actions
/pr-comments GitHub Get PR comments
/review GitHub Review pull requests
AI Enhancement Commands
Command Purpose Level
/think Extended thinking mode Advanced
/upgrade Access Claude Max Premium
/user Personal commands Custom
Experimental Commands
Command Status Purpose
/terminal-setup Proposed Terminal optimizations

Custom Slash Commands

Build Your Own: Create custom commands for repeated workflows

Type Pattern Storage Location Scope
Project /project:<command> .claude/commands/ Current project
Personal /user:<command> ~/.claude/commands/ All projects

Examples:

# Project-specific workflow
/project:deploy     # Deploy this project
/project:test-all   # Run full test suite

# Personal shortcuts
/user:morning       # Daily startup routine
/user:review-pr     # PR review checklist

Environment Variables

Core Configuration Variables

Variable Purpose Example Required
ANTHROPIC_API_KEY API Authentication sk-ant-api03-xxx Yes
ANTHROPIC_MODEL Default model claude-sonnet-4 No
ANTHROPIC_SMALL_FAST_MODEL Quick operations claude-haiku-3 No

Quick Setup:

# Essential - Add to your ~/.bashrc or ~/.zshrc
export ANTHROPIC_API_KEY="sk-ant-api03-your-key-here"

# Optional - Set default model
export ANTHROPIC_MODEL="claude-sonnet-4"

Claude Code Specific Variables

Cloud Provider Integration
Variable Cloud Purpose
CLAUDE_CODE_USE_BEDROCK AWS Enable Bedrock integration
CLAUDE_CODE_USE_VERTEX GCP Enable Vertex AI integration
CLAUDE_CODE_SKIP_BEDROCK_AUTH AWS Skip authentication (gateways)
CLAUDE_CODE_SKIP_VERTEX_AUTH GCP Skip authentication (gateways)
Telemetry & Privacy
Variable Purpose Privacy Impact
CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC Disable all non-essential traffic High Privacy
CLAUDE_CODE_ENABLE_TELEMETRY Control telemetry Configurable

Privacy-First Setup:

# Maximum privacy
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1

Feature Control Variables

Privacy & Security Controls
Variable Feature Privacy Level
DISABLE_AUTOUPDATER=1 Auto-updates Medium
DISABLE_BUG_COMMAND=1 Bug reporting High
DISABLE_ERROR_REPORTING=1 Error reports High
DISABLE_TELEMETRY=1 Analytics High
Cost & Performance Controls
Variable Feature Cost Impact
DISABLE_COST_WARNINGS=1 Cost alerts None
DISABLE_NON_ESSENTIAL_MODEL_CALLS=1 Extra AI calls Saves Money
DISABLE_PROMPT_CACHING=1 Caching Costs More

Recommended Privacy Setup:

# Privacy-focused configuration
export DISABLE_TELEMETRY=1
export DISABLE_ERROR_REPORTING=1
export DISABLE_BUG_COMMAND=1

# Cost optimization
export DISABLE_NON_ESSENTIAL_MODEL_CALLS=1

Network & Proxy Variables

Variable Protocol Example
HTTP_PROXY HTTP http://proxy.company.com:8080
HTTPS_PROXY HTTPS https://proxy.company.com:8443

Corporate Network Setup:

# Corporate proxy configuration
export HTTP_PROXY="http://proxy.company.com:8080"
export HTTPS_PROXY="https://proxy.company.com:8443"

Advanced Configuration Variables

AI Behavior Tuning
Variable Purpose Example
MAX_THINKING_TOKENS Thinking budget MAX_THINKING_TOKENS=50000
MCP_TIMEOUT MCP startup timeout MCP_TIMEOUT=10000
Regional Configuration
Variable Cloud Provider Purpose
AWS_REGION AWS Bedrock us-east-1
CLAUDE_ML_REGION GCP Vertex us-central1

Multi-Region Setup:

# AWS Bedrock in US East
export AWS_REGION="us-east-1"

# GCP Vertex in US Central
export CLAUDE_ML_REGION="us-central1"

Recent Updates (2025)

Version 1.0.25 (Latest)

New Features

  • Fixed slash command reliability issues
  • Improved /mcp output functionality
  • Fixed settings array merge bug

June 18, 2025 Update

  • SSE Transport: Real-time MCP communication
  • HTTP Transport: Web-based MCP servers
  • Remote MCP: Cloud MCP functionality
  • Enhanced Auth: Better MCP authentication

Hidden Features

Secret Sauce: Features most users don't know about

Tool Permission Patterns

Precise Tool Control

Allowlist Patterns

Pattern Scope Example
"Tool" Entire tool "Edit"
"Tool(scope:*)" Scoped access "Bash(git:*)"
"Tool1,Tool2" Multiple tools "Edit,View"
"mcp__server__tool" MCP tools "mcp__puppeteer__navigate"

Common Patterns

# Text editing only
claude --allowedTools "Edit,View"

# Git operations only  
claude --allowedTools "Bash(git:*)"

# Development workflow
claude --allowedTools "Edit,View,Bash(npm:*),Bash(git:*)"

# Web automation
claude --allowedTools "mcp__puppeteer__puppeteer_navigate,mcp__puppeteer__puppeteer_click"

Advanced Usage

Automation Patterns

Headless Automation 
# JSON output for scripts
claude -p "analyze code" --output-format stream-json

# Dangerous mode (containerized only!)
claude --resume abc123 --dangerously-skip-permissions

# Debug everything
claude --verbose --mcp-debug

AI Behavior Control

Prompt Engineering 
# Custom system prompt
claude --system-prompt "You are a security expert" -p "review code"

# Append to system prompt
claude --append-system-prompt "Focus on vulnerabilities" -p "review"

# Thinking levels
claude "think about this"           # Basic (1K tokens)
claude "think hard about this"      # Enhanced
claude "ultrathink about this"      # Maximum (128K tokens)

Interactive Mode Tricks

Keyboard Shortcuts
Shortcut Mode Purpose
Shift+Tab (1x) Auto-Accept Skip confirmations
Shift+Tab (2x) Planning Analyze without writing
Escape Interrupt Stop Claude anytime
Escape (2x) History Edit previous prompts

Workflow Example:

  1. Start Claude Code
  2. Press Shift+Tab twice → Planning mode
  3. Describe task → Claude plans without executing
  4. Review plan → Exit planning mode to execute

Advanced Configurations

Power User Setup 
# Multi-directory workspace
claude --add-dir ../apps ../lib ../shared

# Controlled execution
claude -p --max-turns 3 "refactor function"

# Custom permission handler
claude -p --permission-prompt-tool mcp_auth_tool "write files"

# Specific model selection
claude --model claude-sonnet-4-20250514 "analyze codebase"

# Extended thinking + tools (beta)
claude "ultrathink about this migration and use tools as needed"

Power User Aliases

Community Shortcuts

⚠️ Warning: Use dangerous aliases only in safe environments!

# Quick access (add to ~/.bashrc or ~/.zshrc)
alias cc="claude --dangerously-skip-permissions"  # DANGEROUS
alias yolo="claude --dangerously-skip-permissions" # VERY DANGEROUS

# Output formats
alias claude-json="claude -p --output-format json"
alias claude-stream="claude -p --output-format stream-json"

# Safe shortcuts
alias cl="claude"                                   # Simple shortcut
alias clp="claude -p"                              # Print mode
alias cls="claude /status"                         # Quick status

Configuration Files

Settings Files

File Scope Purpose
~/.claude.json Global User-wide settings
.claude/settings.json Project Project-specific config
~/.config/claude-code/auth.json Auth Authentication data

Custom Commands

Directory Scope Command Pattern
.claude/commands/ Project /project:command
~/.claude/commands/ Personal /user:command

MCP Configuration

File Purpose
.claude/mcp.json MCP server config
~/.claude-code-mcp.env MCP environment vars

Security

Dangerous Flags

⚠️ EXTREME CAUTION REQUIRED

Flag Risk Level Impact
--dangerously-skip-permissions CRITICAL Bypasses ALL security checks

Safe Usage Guidelines:

  • Only use in containerized environments
  • Never use with internet access
  • Can cause data loss/corruption
  • Major security vulnerability

Safe Usage Patterns

Security Best Practices

Always Review First

  • Review commands before granting permissions
  • Use /permissions to audit settings
  • Use /doctor for security health checks

Principle of Least Privilege

  • Avoid blanket permissions
  • Use specific tool allowlists
  • Regular permission audits

Security Commands

# Check security status
claude /doctor

# Review permissions
claude /permissions

# Safe tool allowlist
claude --allowedTools "Edit,View"

Output Formats

Text Output (Default)

# Standard human-readable output
claude -p "explain this function" --output-format text
Example Output 
This function calculates the factorial of a number recursively...

JSON Output

# Structured data with metadata
claude -p "analyze this code" --output-format json
Example Output 
{
  "response": "This function calculates...",
  "metadata": {
    "cost": 0.0023,
    "duration": 1.2,
    "model": "claude-sonnet-4"
  }
}

Streaming JSON Output

# Real-time streaming for large operations
claude -p "process large file" --output-format stream-json
Example Output 
{"type": "start", "timestamp": "2025-06-23T10:00:00Z"}
{"type": "progress", "content": "Processing..."}
{"type": "result", "content": "Analysis complete"}
{"type": "end", "timestamp": "2025-06-23T10:00:05Z"}

⚠️ Note: Each line is valid JSON, but concatenated output is not valid JSON

Integration Examples

CI/CD Integration

Build Pipeline Examples 
# Automated fixing in build scripts
claude -p "fix linting errors" --dangerously-skip-permissions --output-format json

# Controlled tool access
claude -p "run tests and fix failures" --allowedTools "Bash(npm:*),Edit"

# Pipeline integration
if claude -p "check code quality" --output-format json | jq -r '.passed'; then
  echo "Quality checks passed"
else
  echo "Quality issues found"
  exit 1
fi

Scripting Integration

Automation Scripts 
# Log analysis
cat error.log | claude -p "summarize errors" --output-format json > summary.json

# Batch processing
for file in *.py; do
    claude -p "review this Python file" < "$file" --output-format text >> review.txt
done

# Monitoring integration
journal_errors=$(journalctl --since "1 hour ago" --priority=err)
echo "$journal_errors" | claude -p "analyze system errors" --output-format json > /tmp/error_analysis.json

Troubleshooting

Health Checks

Command Purpose When to Use
claude --help 📚 Show all options Getting started
/doctor 🩺 Health check Something's wrong
/status 📊 Current config Check settings
claude config list 📝 List all settings Audit configuration

Debug Mode

Debug Commands 
# Detailed logging
claude --verbose

# Full debug mode
claude --debug

# MCP-specific debugging
claude --mcp-debug

# Combined debugging
claude --verbose --debug --mcp-debug

Reset & Recovery

Emergency Recovery 
# Reset authentication
rm -rf ~/.config/claude-code/auth.json

# Clean logout/login
claude /logout
claude /login

# Reset all settings
rm -rf ~/.claude.json
rm -rf .claude/settings.json

# Full health check after reset
claude /doctor

Beta Features

Interactive Mode Controls

Shortcut Action Mode
Shift+Tab (1x) Auto-Accept Skip confirmations
Shift+Tab (2x) Planning Analyze without writing
Escape Interrupt Stop Claude anytime
Escape (2x) History Edit previous prompts

Thinking Budget Control

Secret: Special phrases control AI thinking depth

Thinking Levels
Phrase Thinking Budget Use Case
"think" 1,024 tokens Basic analysis
"think hard" Enhanced Complex problems
"think harder" Advanced Deep analysis
"think more" Alternative enhanced Thorough review
"ultrathink" 128K tokens Maximum depth
"megathink" Alternative max Alternative maximum

Examples:

# Basic thinking
claude "think about this algorithm"

# Enhanced thinking
claude "think hard about the security implications"

# Maximum thinking
claude "ultrathink about this database migration"

Extended Thinking Mode

Advanced Thinking Controls 
# Slash command trigger
/think

# Custom budget (proposed)
--thinking-budget 50000

Permission Prompt Tool Integration

Custom Permission Handler

The --permission-prompt-tool expects an MCP tool that returns JSON:

Allow Execution

{
  "behavior": "allow",
  "updatedInput": {...}  // original or modified input
}

Deny Execution

{
  "behavior": "deny",
  "message": "Reason for denial"
}

GitHub Integration Features

Feature Status Description
@claude mentions Beta PR/Issue mentions
GitHub Actions Beta Automated workflows
gh CLI integration Stable Repository operations
IDE extensions Beta VS Code & JetBrains
Inline edits Beta IDE interface edits

Experimental Features

Cutting-Edge Capabilities

AI Enhancements

  • Extended thinking with tool use: Claude uses tools during thinking
  • Parallel tool execution: Multiple tools run simultaneously
  • 128K output tokens: 15x longer responses

Interactive Modes

  • Planning mode: Shift+Tab twice for analysis without writes
  • Auto-accept mode: Shift+Tab once for autonomous implementation

Integrations

  • IDE integrations: VS Code and JetBrains extensions
  • Advanced memory system: Multi-tiered contextual memory

Important Notes

Command Precedence & Behavior

Command Precedence

Query > Session > App Config > Environment Variables > Defaults

Key Behaviors

  1. Session Management: Each command can specify session name for isolation
  2. Tool Names: Use exact names like "Replace", "Edit", "Bash", "View"
  3. MCP Integration: Model Context Protocol with SSE/HTTP support
  4. Version Updates: Use claude update for latest features
  5. DMCA Notice: Decompiled source repository was taken down
  6. Beta Features: GitHub Actions, IDE extensions in beta
  7. Thinking Modes: Special phrases trigger different thinking levels
  8. Latest Version: 1.0.25 (June 2025) with improved reliability
  9. Real-time Features: SSE transport enables real-time MCP communication

✅ Research Verification

COMPLETE RESEARCH CONDUCTED

Source Status
Official Documentation Thoroughly Reviewed
GitHub Repository & Issues Comprehensively Analyzed
Community Discussions Extensively Searched
CLI Reference Completely Documented
Environment Variables Fully Catalogued
Slash Commands Entirely Mapped
Hidden Features Discovered & Documented
2025 Updates Latest Changelog Reviewed
Source Code Analysis Completed (pre-DMCA)

The Most Complete Claude Code Reference

This comprehensive guide covers every discoverable Claude Code command as of June 2025,
including many features that are not widely known or documented in basic tutorials.

This represents the most complete Claude Code command reference available.

For updates and contributions, visit the official Claude Code documentation

Reference in New Issue View Git Blame Copy Permalink