Claude Code Tooling Index: Make Your Agent Stack Searchable

The Problem

Once you start building agent tooling seriously, you hit “tooling drift”:

• Skills and plugins multiply
• Hooks evolve across repos
• MCP servers come and go
• Nobody remembers what’s installed, what’s used, or what’s broken

The result is a painful failure mode: you have powerful tools, but no map.

The Solution

Claude Code Tooling Index turns your setup into a searchable, measurable system:

• Scans your Claude Code environment
• Normalizes metadata across component types
• Stores it locally (SQLite + full-text search)
• Exposes it via CLI + a fast TUI

Architecture

The scanner is an orchestrator that fans out to specialized scanners:

tooling-index scan
  ├─ skills scanner
  ├─ plugins scanner
  ├─ commands scanner
  ├─ hooks scanner
  ├─ MCP config scanner
  └─ binaries scanner
        ↓
     SQLite + FTS5
        ↓
   tooling-index tui

The dashboard (TUI) is optimized for the workflow that actually matters:

“I know I have a tool for this… what’s it called and how do I invoke it?”

Why Local-First

Tooling inventories often reveal:

• your habits
• your workflow patterns
• what you’re building privately

Local-first keeps the index useful without creating a new privacy liability.

What I’d Add Next

• First-class “broken / stale / unused” health scoring
• One-command export to a shareable “tooling snapshot”
• A lightweight “recommendations” layer (e.g., “you installed X but never used it”)