create-claudify

Scaffold a production-grade Claude Code systems architecture into any project.

Turn Claude Code from a chatbot into an autonomous operating system with memory, hooks, commands, agents, and skills.

Quick Start

npx create-claudify

Answer the prompts. That's it.

What You Get

your-project/
├── CLAUDE.md                    # Project instructions — Claude reads this first
├── CLAUDE.local.md              # Personal overrides (gitignored)
├── SETUP.md                     # Getting started guide
├── .mcp.json                    # MCP server config
└── .claude/
    ├── settings.json            # Permissions + hooks
    ├── memory.md                # Living session state
    ├── knowledge-base.md        # Institutional memory
    ├── knowledge-nominations.md # Candidate learnings pipeline
    ├── command-index.md         # Command registry
    ├── commands/                # /start, /sync, /clear, /wrap-up, /audit
    ├── agents/                  # auditor + reviewer
    ├── agent-memory/            # Persistent agent knowledge
    ├── skills/                  # research, deploy, monitor, data-export, content
    ├── hooks/                   # quality-gate, context-monitor, audit-logger
    └── logs/                    # Audit trail

Features

Choose what to include during setup:

| System | What it does | |---|---| | Memory | 4-layer architecture: global, project, personal, dynamic. Claude remembers across sessions. | | Commands | Daily workflow rituals: /start, /sync, /clear, /wrap-up, /audit. | | Hooks | Deterministic enforcement: block bad writes, monitor context usage, log all tool calls. | | Agents | Specialist subagents: auditor (quality gate) + reviewer (read-only QA). | | Skills | Domain knowledge loaded on demand: research, deploy, monitor, data-export, content. | | MCP | Pre-configured MCP servers: context7 (live docs) + memory (knowledge graph). |

Update

When we release new templates, commands, or improvements:

npx create-claudify update

This updates system files (hooks, commands, agents, skills) without touching your custom content (CLAUDE.md, memory.md, knowledge-base.md, settings.json).

How It Works

1. CLAUDE.md is the routing file — Claude reads it first, follows links to everything else
2. Memory system gives Claude session state + institutional knowledge
3. Hooks enforce rules deterministically (no relying on instructions alone)
4. Commands are repeatable procedures triggered by name
5. Agents are specialist subprocesses with their own memory
6. Skills are domain knowledge loaded only when needed

Requirements

• Node.js 18+
• Claude Code installed

License

MIT