System Architecture

The Claude Code Development Kit is built around an Integrated Intelligence Loop that connects three core components: Commands, Documentation, and MCP Servers. This architecture ensures that Claude Code always operates with the most current and relevant context.

Integrated Intelligence Loop

                        CLAUDE CODE
                   ┌─────────────────┐
                   │                 │
                   │    COMMANDS      │
                   │                 │
                   └────────┬────────┘
                  Multi-agent│orchestration
                   Parallel │execution
                   Dynamic  │scaling
                           ╱│╲
                          ╱ │ ╲
          Routes agents  ╱  │  ╲  Leverages
          to right docs ╱   │   ╲ expertise
                       ╱    │    ╲
                      ▼     │     ▼
         ┌─────────────────┐│┌─────────────────┐
         │                 │││                 │
         │  DOCUMENTATION  │││  MCP SERVERS   │
         │                 │││                 │
         └─────────────────┘│└─────────────────┘
          3-tier structure  │  Context7 + Gemini
          Auto-loading      │  Real-time updates
          Context routing   │  AI consultation
                      ╲     │     ╱
                       ╲    │    ╱
        Provides project╲   │   ╱ Enhances with
        context for      ╲  │  ╱  current best
        consultation      ╲ │ ╱   practices
                           ╲│╱
                            ▼
                    Integrated Workflow

Auto-Loading Mechanism

A key feature of the kit is the automatic loading of critical documentation for every command execution. The following files are always included:

@/CLAUDE.md                              # Master AI context and coding standards
@/docs/ai-context/project-structure.md   # Complete technology stack and file tree
@/docs/ai-context/docs-overview.md       # Documentation routing map

This mechanism is extended to all sub-agents via the subagent-context-injector.sh hook. This ensures: - Consistent AI behavior across all sessions and sub-agents. - Zero manual context management is required at any level.

Component Integration

The three core components work together seamlessly:

• Commands ↔️ Documentation: Commands use the documentation structure to determine which context to load and how to orchestrate sub-agents. They also update the documentation to keep it synchronized with the codebase.
• Commands ↔️ MCP Servers: Commands leverage MCP servers like Context7 for up-to-date library information and Gemini for architectural consultations, integrating this external expertise directly into workflows.
• Documentation ↔️ MCP Servers: Key documentation files, like the project structure and assistant rules, are automatically attached to Gemini consultations. This ensures external AI advice is relevant and compliant with your project's standards.

Hooks Integration

The kit includes a set of battle-tested hooks that enhance Claude Code's native capabilities:

• Security Scanner: Prevents accidental exposure of secrets when using MCP servers.
• Gemini Context Injector: Automatically includes project structure in Gemini consultations.
• Subagent Context Injector: Ensures all sub-agents receive core documentation automatically.
• Notification System: Provides optional, non-blocking audio feedback for task completion and input requests.

These hooks integrate seamlessly into the workflow, providing pre-execution security checks, automatic context enhancement, and an improved developer experience.