Claude Code Configuration Guide

Help users configure Claude Code for optimal development workflows.

Quick Reference

CLAUDE.md Structure

The CLAUDE.md file provides project context to Claude. Keep it concise (<500 lines).

# Project Name

Brief description of what this project does.

## Architecture

Key directories, patterns, and data flow.

## Development Rules

### ALWAYS
- Critical patterns to follow

### NEVER
- Anti-patterns to avoid

## Quick Reference

| Command | Purpose |
|---------|---------|
| `npm run dev` | Start development |
| `npm test` | Run tests |

## Key Files

| File | Purpose |
|------|---------|
| `src/index.ts` | Entry point |

Best practices:

• Focus on what Claude needs to know, not general documentation
• Include commands Claude should run
• List critical constraints and anti-patterns
• Use tables for quick reference

For detailed CLAUDE.md patterns: See references/claude-md.md

Plugin Architecture

Plugins package commands, skills, hooks, and MCP servers together.

plugins/my-plugin/
├── plugin.json           # Metadata (required)
├── commands/             # Slash commands
│   └── my-command.md
├── skills/               # Agent skills
│   └── my-skill/
│       └── SKILL.md
├── agents/               # Subagent definitions
│   └── my-agent.md
└── hooks/                # Hook scripts
    └── pre-tool.sh

plugin.json:

{
  "name": "my-plugin",
  "version": "1.0.0",
  "description": "Plugin description"
}

For plugin development guide: See references/plugin-structure.md

Skill Patterns

Skills extend Claude's capabilities with domain-specific knowledge.

.claude/skills/my-skill/
├── SKILL.md              # Instructions (required)
├── references/           # Detailed docs (loaded on demand)
├── scripts/              # Executable code
└── assets/               # Templates

SKILL.md structure:

---
name: my-skill
description: "When to trigger this skill"
---

# Skill Name

Brief description.

## When to Use

Specific activation criteria.

## Instructions

Step-by-step guidance.

## Quick Reference

For detailed X: See `references/x.md`

Progressive disclosure: Keep SKILL.md concise (<200 lines). Put detailed documentation in references/ directory.

For skill development patterns: See references/skills.md

Hooks Configuration

Hooks execute shell commands in response to Claude Code events.

Configure in .claude/settings.local.json:

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": ["./scripts/validate-bash.sh"]
      }
    ],
    "PostToolUse": [
      {
        "matcher": "Write",
        "hooks": ["prettier --write $CLAUDE_FILE_PATH"]
      }
    ],
    "UserPromptSubmit": [
      {
        "hooks": ["./scripts/log-prompt.sh"]
      }
    ]
  }
}

Hook types:

• PreToolUse - Before tool execution (can block)
• PostToolUse - After tool execution
• UserPromptSubmit - When user submits prompt

Environment variables available:

• $CLAUDE_TOOL_NAME - Tool being called
• $CLAUDE_FILE_PATH - File path (for file tools)

For hook patterns and examples: See references/hooks.md

MCP Integration

Model Context Protocol connects Claude to external tools and services.

Configure in .mcp.json:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/allowed/path"]
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
      }
    }
  }
}

Common MCP servers:

• @modelcontextprotocol/server-filesystem - File access
• @modelcontextprotocol/server-github - GitHub integration
• @modelcontextprotocol/server-postgres - Database queries
• @modelcontextprotocol/server-puppeteer - Browser automation

For MCP configuration guide: See references/mcp.md

Settings Configuration

Project settings (.claude/settings.json):

{
  "model": "sonnet",
  "maxTokens": 8192,
  "thinking": {
    "enabled": true,
    "budget": 10000
  }
}

Model aliases:

• sonnet - Balanced (default)
• opus - Complex tasks
• haiku - Fast, simple tasks

For full settings reference: See references/settings.md

Slash Commands

Create custom commands in .claude/commands/:

.claude/commands/deploy.md:

---
description: Deploy to production
---

Run the deployment pipeline:
1. Run tests: `npm test`
2. Build: `npm run build`
3. Deploy: `./scripts/deploy.sh`

Verify deployment succeeded and report status.

Invoke with: /deploy

Arguments: Use $ARGUMENTS for user input

---
description: Create a component
---

Create a React component named $ARGUMENTS in src/components/.

For command patterns: See references/commands.md

Best Practices

Project Setup Checklist

1. CLAUDE.md - Project context and rules
2. .claude/settings.json - Model and token settings
3. .mcp.json - External tool integrations
4. .claude/commands/ - Frequently used workflows
5. .claude/skills/ - Domain-specific knowledge

Common Patterns

Auto-format on write:

{
  "hooks": {
    "PostToolUse": [
      {"matcher": "Write", "hooks": ["prettier --write $CLAUDE_FILE_PATH"]}
    ]
  }
}

Validate before bash:

{
  "hooks": {
    "PreToolUse": [
      {"matcher": "Bash", "hooks": ["./scripts/validate-command.sh"]}
    ]
  }
}

Security

• Never commit API keys in .mcp.json - use ${ENV_VAR} syntax
• Use hooks to validate dangerous operations
• Restrict MCP filesystem access to project directories

Reference Files

Load these for detailed guidance:

Codex vs Claude Code Comparison