Essential Configuration

Claude Code’s flexibility comes from its extensive configuration options. This guide covers essential settings to optimize your development workflow, from permission management to model selection strategies.

Configuration Hierarchy

1. Enterprise policies - /Library/Application Support/ClaudeCode/managed-settings.json (macOS)
2. Command line arguments - Flags passed when launching Claude Code
3. Local project settings - .claude/settings.local.json (not in source control)
4. Shared project settings - .claude/settings.json (checked into source control)
5. User settings - ~/.claude/settings.json (applies to all projects)

Settings are applied in order of precedence, with enterprise policies overriding all others.

Quick Configuration Setup

Interactive

# Launch configuration menu
/config

# View current settings
claude config list

# Set a specific option
claude config set model claude-sonnet-4

JSON Configuration

.claude/settings.json

{
  "model": "claude-sonnet-4",
  "permissions": {
    "defaultMode": "acceptEdits",
    "allow": [
      "Edit",
      "Bash(npm test:*)",
      "Bash(git:*)"
    ]
  },
  "hooks": {
    "PostToolUse": {
      "Edit": "npm run lint --fix $CLAUDE_FILE_PATHS"
    }
  }
}

Permission Modes

Claude Code offers several permission modes to balance safety and productivity:

Available Modes

Mode	Description	Use Case
default	Prompts for each new tool use	General development
acceptEdits	Auto-accepts file edits for session	Active coding sessions
plan	Read-only analysis mode	Code review, architecture planning
bypassPermissions	Skip all permission prompts	Trusted environments only

Yolo Mode (Bypass Permissions)

Caution

Only use in trusted environments like local development or isolated containers.

# Launch with permissions bypassed
claude --dangerously-skip-permissions

# Equivalent to setting in config
claude --permission-mode bypassPermissions

This mode eliminates interruptions from permission prompts, significantly improving workflow speed for experienced users.

Model Selection Strategy

Claude Code defaults to intelligent model switching, but you can optimize for your use case:

Model Capabilities

Model	Best For	Cost	Speed
Claude Opus 4	Complex reasoning, architecture	5x	Slower
Claude Sonnet 4	General coding, implementation	1x	Fast
OpenAI o3	Debugging complex issues	Varies	Medium
Gemini 2.5 Pro	Large context scenarios	2x	Fast

Dynamic Model Selection

# Switch models mid-session
/model opus     # For complex planning
/model sonnet   # For implementation

# Set default model
claude config set model claude-sonnet-4

# Override per session
claude --model claude-opus-4

Cost-Optimized Configuration

{
  "model": "claude-sonnet-4",
  "permissions": {
    "defaultMode": "acceptEdits"
  },
  "env": {
    "DISABLE_NON_ESSENTIAL_MODEL_CALLS": "1",
    "DISABLE_COST_WARNINGS": "0"
  }
}

Project-Level Configuration

Create .claude/settings.json for team-shared settings:

{
  "permissions": {
    "allow": [
      "Edit(src/**)",
      "Bash(npm run:*)",
      "Bash(git diff:*)",
      "mcp__github__*"
    ],
    "deny": [
      "Edit(*.env)",
      "Bash(rm -rf:*)"
    ],
    "additionalDirectories": ["../shared-lib"]
  },
  "env": {
    "NODE_ENV": "development"
  },
  "hooks": {
    "PostToolUse": {
      "Edit": "prettier --write $CLAUDE_FILE_PATHS"
    }
  }
}

IDE Integration Settings

VS Code

.vscode/settings.json

{
  "claude-code.autoLaunch": true,
  "claude-code.defaultMode": "acceptEdits",
  "claude-code.theme": "dark"
}

JetBrains

.idea/claude-code.xml

<component name="ClaudeCodeSettings">
  <option name="autoLaunch" value="true" />
  <option name="defaultMode" value="acceptEdits" />
</component>

Terminal Configuration

Optimize your terminal for Claude Code:

1. Run /terminal-setup to configure key bindings
2. Enable Shift+Enter for multiline input
3. Configure notification preferences:

   claude config set -g preferredNotifChannel iterm2_with_bell

Memory Configuration (CLAUDE.md)

Configure persistent memory across sessions:

CLAUDE.md

## Build Commands
- npm run build: Production build
- npm run dev: Development server
- npm test: Run test suite
- npm run typecheck: TypeScript validation

## Code Style
- Use functional components with TypeScript
- Prefer composition over inheritance
- Always handle error cases explicitly

## Architecture
- Frontend: Next.js with App Router
- API: tRPC with Zod validation
- Database: PostgreSQL with Drizzle ORM

Memory File Imports

# Project-specific settings
@docs/conventions.md
@~/.claude/personal-preferences.md

# Team member preferences (not in git)
@~/.claude/team-configs.md

MCP Server Configuration

Enable powerful integrations through MCP:

.claude/mcp.json

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
      }
    },
    "postgres": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-postgres",
        "postgresql://localhost/mydb"
      ]
    }
  }
}

Enable MCP servers:

# Interactive setup
claude mcp add

# Allow specific servers
claude config set enabledMcpjsonServers '["github", "postgres"]'

Advanced Hook Configuration

Automate workflows with hooks:

{
  "hooks": {
    "PreToolUse": {
      "Edit": {
        "type": "command",
        "command": "echo 'Editing: $CLAUDE_FILE_PATHS'"
      }
    },
    "PostToolUse": {
      "Edit": {
        "type": "command",
        "command": "npm run lint --fix $CLAUDE_FILE_PATHS"
      },
      "Bash": {
        "type": "command",
        "command": "[[ \"$CLAUDE_TOOL_INPUT\" =~ test ]] && echo 'Tests complete!'"
      }
    }
  }
}

Environment Variables

Key environment variables for fine-tuning:

# Model configuration
export ANTHROPIC_MODEL='claude-opus-4'
export ANTHROPIC_SMALL_FAST_MODEL='claude-3-5-haiku'

# Performance tuning
export CLAUDE_CODE_MAX_OUTPUT_TOKENS=8192
export MAX_THINKING_TOKENS=50000

# Cost optimization
export DISABLE_NON_ESSENTIAL_MODEL_CALLS=1

# Security
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1

Custom Slash Commands

Create reusable workflows:

.claude/commands/optimize.md

Analyze this code for performance bottlenecks:
- Profile execution time
- Identify N+1 queries
- Check for unnecessary re-renders
- Suggest optimizations

Target: $ARGUMENTS

Usage:

/optimize src/components/Dashboard.tsx

Security Configuration

Enterprise Security Settings

/etc/claude-code/managed-settings.json

{
  "permissions": {
    "disableBypassPermissionsMode": "disable",
    "deny": [
      "Bash(curl:*)",
      "Bash(wget:*)",
      "WebFetch"
    ]
  },
  "forceLoginMethod": "console"
}

Performance Optimization

Token Usage

• Use /clear frequently
• Enable auto-compact
• Focus queries on specific modules

Model Selection

• Sonnet 4 for routine tasks
• Opus 4 for complex reasoning
• Monitor with /cost command

Quick Reference

Essential Commands

/config          # Configuration menu
/permissions     # Manage allowed tools
/model          # Switch models
/clear          # Clear context
/cost           # Check token usage

Common Configurations

High-Speed Development:

claude --dangerously-skip-permissions --model sonnet

Secure Review Mode:

claude --permission-mode plan --model opus

Cost-Conscious Setup:

{
  "model": "claude-sonnet-4",
  "env": {
    "DISABLE_NON_ESSENTIAL_MODEL_CALLS": "1"
  }
}

Next Steps

• Set up project initialization with CLAUDE.md
• Configure MCP servers for extended capabilities
• Learn about deep reasoning modes for complex tasks