toonify-mcp 0.1.0 → 0.1.1

package/README.md

@@ -1,123 +1,60 @@
-# 🎯 Claude Code Toonify
+# 🎯 Toonify MCP
 
 **[English](README.md) | [繁體中文](README.zh-TW.md)**
 
-**Reduce Claude API token usage by 60%+ using TOON format optimization**
+MCP server providing token optimization tools for converting structured data to TOON format.
+Reduces Claude API token usage by 60%+ on average.
 
-An MCP (Model Context Protocol) server that provides **on-demand** token optimization tools for converting structured data to TOON (Token-Oriented Object Notation) format. Works with any MCP-compatible LLM client (Claude Code, ChatGPT, etc.).
+## Features
 
-⚠️ **Important**: This MCP server provides **tools** that must be explicitly called - it does NOT automatically intercept content. See [Usage](#-usage) for details.
+- **60%+ Token Reduction** for JSON, CSV, YAML data
+- **MCP Integration** - Works with Claude Code, Claude Desktop
+- **Built-in Metrics** - Track token savings locally
+- **Silent Fallback** - Never breaks your workflow
 
-## 🌟 Features
-
-- **🎯 60%+ Token Reduction**: Average 63.9% reduction for structured data
-- **💰 Significant Cost Savings**: ~$1,380 per million API calls
-- **⚡ Fast**: <10ms overhead for typical payloads
-- **🔄 Silent Fallback**: Never breaks your workflow
-- **📊 Built-in Metrics**: Track savings locally
-- **🔌 MCP Integration**: Works seamlessly with Claude Code
-- **🌍 Universal Compatibility**: Works with ANY LLM (GPT, Claude, Gemini, etc.)
-
-## 📊 Performance
-
-| Format | Before | After | Savings |
-|--------|--------|-------|---------|
-| JSON   | 247 bytes | 98 bytes | **60%** |
-| CSV    | 180 bytes | 65 bytes | **64%** |
-| YAML   | 215 bytes | 89 bytes | **59%** |
-
-**Token reduction example:**
-```
-JSON (142 tokens):
-{
-  "products": [
-    {"id": 101, "name": "Laptop Pro", "price": 1299},
-    {"id": 102, "name": "Magic Mouse", "price": 79}
-  ]
-}
-
-TOON (57 tokens, 60% reduction):
-products[2]{id,name,price}:
-  101,Laptop Pro,1299
-  102,Magic Mouse,79
-```
-
-## 🚀 Installation
+## Installation
 
 ### 1. Install the package
 
 ```bash
-npm install -g @ktseng/toonify-mcp
+npm install -g toonify-mcp
 ```
 
 ### 2. Register with Claude Code
 
 ```bash
-# Register the MCP server (user scope - available in all projects)
+# User scope (available in all projects)
 claude mcp add --scope user --transport stdio toonify -- /opt/homebrew/bin/toonify-mcp
 
-# For project-specific registration
+# Or project scope
 claude mcp add --scope project --transport stdio toonify -- /opt/homebrew/bin/toonify-mcp
 ```
 
 ### 3. Verify installation
 
 ```bash
-# Check if MCP server is registered and connected
 claude mcp list
-
 # Should show: toonify: /opt/homebrew/bin/toonify-mcp - ✓ Connected
 ```
 
-## 📖 Usage
+## Usage
 
-### How It Works: Tool-Based Optimization
+### Option A: MCP Tools (Manual)
 
-This MCP server provides **two tools** that the LLM can call:
-1. `optimize_content` - Optimizes structured data to TOON format
-2. `get_stats` - Returns optimization statistics
+Call tools explicitly when needed:
 
-⚠️ **Key Limitation**: The LLM must **explicitly decide** to call these tools. There is NO automatic interception of content.
-
-### Usage Patterns
-
-**Pattern 1: Explicit User Request**
-```
-User: "Optimize this JSON data for me"
-LLM: [Calls optimize_content tool] → Returns optimized TOON format
-```
-
-**Pattern 2: LLM Decides to Optimize**
-```
-LLM reads large JSON → Recognizes it's large → Calls optimize_content → Uses optimized version
-```
+```bash
+# Optimize content
+claude mcp call toonify optimize_content '{"content": "..."}'
 
-**Pattern 3: Custom Instructions** (ChatGPT/Claude)
-```
-"Before analyzing large JSON/CSV/YAML data, always call the optimize_content tool to reduce tokens"
+# View stats
+claude mcp call toonify get_stats '{}'
 ```
 
-### For ChatGPT Users
+### Option B: Claude Code Hook (Automatic) ⭐ RECOMMENDED
 
-ChatGPT can use this MCP server, but:
-- ❌ **NOT automatic** - ChatGPT must decide to call the tool
-- ✅ **Works when prompted** - "Use toonify to optimize this data"
-- ✅ **Custom instructions** - Add to ChatGPT custom instructions to always optimize large content
+Automatic interception for Claude Code users:
 
-### For Claude Code Users
-
-#### Option A: MCP Server (Manual)
-- ❌ **NOT automatic** - Claude must decide to call the tool
-- ✅ **Works when prompted** - "Use toonify to optimize this data"
-- ✅ **Universal compatibility** - Works with any MCP client
-
-#### Option B: Claude Code Hook (Automatic) ⭐ RECOMMENDED
-- ✅ **Fully automatic** - Intercepts tool results transparently
-- ✅ **Zero overhead** - No manual calls needed
-- ✅ **Seamless integration** - Works with Read, Grep, and other file tools
-- ⚠️ **Claude Code only** - Doesn't work with other MCP clients
-
-**Installation**:
 ```bash
 cd hooks/
 npm install
@@ -128,11 +65,9 @@ npm run install-hook
 claude hooks list  # Should show: PostToolUse
 ```
 
-See `hooks/README.md` for detailed setup and configuration.
+See `hooks/README.md` for details.
 
-## 🔧 Configuration
-
-Configure via environment variables or config file:
+## Configuration
 
 ```bash
 # Environment variables
@@ -141,7 +76,7 @@ export TOONIFY_MIN_TOKENS=50
 export TOONIFY_MIN_SAVINGS=30
 export TOONIFY_SKIP_TOOLS="Bash,Write"
 
-# Or via ~/.claude/toonify-config.json
+# Or ~/.claude/toonify-config.json
 {
   "enabled": true,
   "minTokensThreshold": 50,
@@ -150,195 +85,12 @@ export TOONIFY_SKIP_TOOLS="Bash,Write"
 }
 ```
 
-## 📊 Metrics Dashboard
-
-View your savings:
-
-```bash
-claude mcp call toonify get_stats '{}'
-```
-
-Output:
-```
-📊 Token Optimization Stats
-━━━━━━━━━━━━━━━━━━━━━━━━
-Total Requests: 1,234
-Optimized: 856 (69.4%)
-
-Tokens Before: 1,456,789
-Tokens After: 567,234
-Total Savings: 889,555 (61.1%)
-
-💰 Cost Savings (at $3/1M input tokens):
-   $2.67 saved
-```
-
-## 🌍 Compatibility
-
-### ✅ **This MCP Server Works With:**
-- **Claude Code CLI** (primary target)
-- **Claude Desktop App**
-- **Custom MCP clients**
-- **Any tool implementing MCP protocol**
-
-**Important**: MCP (Model Context Protocol) is an Anthropic protocol. This MCP server only works with MCP-compatible clients in the Claude ecosystem.
-
-### 🔧 **Using TOON Format with Other LLMs**
-
-While this **MCP server** is Claude-specific, the **TOON format itself** reduces tokens for ANY LLM (GPT, Gemini, Llama, etc.). To use TOON optimization with non-MCP LLMs:
-
-**TypeScript/JavaScript:**
-```typescript
-import { encode, decode } from '@toon-format/toon';
-
-// Optimize data before sending to any LLM API
-const data = {
-  products: [
-    { id: 101, name: 'Laptop Pro', price: 1299 },
-    { id: 102, name: 'Magic Mouse', price: 79 }
-  ]
-};
-
-const optimizedContent = encode(data); // 60% token reduction
-
-// Use with OpenAI
-await openai.chat.completions.create({
-  model: 'gpt-4',
-  messages: [{ role: 'user', content: `Analyze: ${optimizedContent}` }]
-});
-
-// Use with Gemini
-await gemini.generateContent({
-  contents: [{ text: `Analyze: ${optimizedContent}` }]
-});
-```
-
-**Python:**
-```python
-# Install: pip install toonify
-from toonify import encode
-import openai
-
-data = {"products": [...]}
-optimized = encode(data)
-
-# Works with any LLM API
-openai.chat.completions.create(
-    model="gpt-4",
-    messages=[{"role": "user", "content": f"Analyze: {optimized}"}]
-)
-```
-
-### 📊 **MCP Server vs TOON Library**
-
-| Feature | This MCP Server | TOON Library Direct |
-|---------|----------------|---------------------|
-| **Target** | Claude Code/Desktop | Any LLM |
-| **Integration** | Automatic (via MCP) | Manual (code integration) |
-| **Setup** | Configure once | Import in each project |
-| **Compatibility** | MCP clients only | Universal |
-
-## 🏗️ Architecture
-
-```
-┌──────────────────┐
-│   Claude Code    │
-│   CLI            │
-└─────────┬────────┘
-          │
-          ↓
-┌──────────────────┐
-│  Toonify MCP     │
-│  Server          │
-├──────────────────┤
-│ • optimize_content│ ← Compress structured data
-│ • get_stats      │ ← View metrics
-└─────────┬────────┘
-          │
-          ↓
-┌──────────────────┐
-│ TokenOptimizer   │ ← TOON encoding
-└──────────────────┘
-          │
-          ↓
-┌──────────────────┐
-│ MetricsCollector │ ← Track savings
-└──────────────────┘
-```
-
-## 🧪 Development
-
-```bash
-# Clone repository
-git clone https://github.com/ktseng/toonify-mcp.git
-cd toonify-mcp
-
-# Install dependencies
-npm install
-
-# Build
-npm run build
-
-# Run tests
-npm test
-
-# Development mode (watch)
-npm run dev
-```
-
-## 🎯 When to Use
-
-**✅ Ideal for:**
-- Large JSON responses from tools
-- CSV/tabular data
-- Structured API responses
-- Database query results
-- File contents with structured data
-
-**⚠️ Not recommended for:**
-- Plain text content
-- Code files
-- Highly irregular data structures
-- Content < 50 tokens
-
-## 📚 Technical Details
-
-### How It Works
-
-1. **Interception**: MCP server intercepts tool results via `optimize_content` tool
-2. **Detection**: Analyzes content to identify JSON, CSV, or YAML
-3. **Optimization**: Converts to TOON format using `@scrapegraph/toonify`
-4. **Validation**: Ensures savings > 30% threshold
-5. **Fallback**: Returns original if optimization fails or savings too low
-
-### Dependencies
-
-- `@modelcontextprotocol/sdk` - MCP server framework
-- `@scrapegraph/toonify` - TOON format encoder/decoder
-- `tiktoken` - Token counting (Claude/GPT compatible)
-- `yaml` - YAML parsing support
-
-## 🤝 Contributing
-
-Contributions welcome! Please see [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
-
-## 📄 License
-
-MIT License - see [LICENSE](LICENSE) file
-
-## 🙏 Credits
-
-- [Toonify](https://github.com/ScrapeGraphAI/toonify) by ScrapeGraphAI team
-- [Claude Code](https://github.com/anthropics/claude-code) by Anthropic
-- Original inspiration: [awesome-llm-apps](https://github.com/Shubhamsaboo/awesome-llm-apps)
-
-## 🔗 Links
+## Links
 
-- **NPM Package**: Coming soon
-- **GitHub**: https://github.com/ktseng/toonify-mcp
-- **Issues**: https://github.com/ktseng/toonify-mcp/issues
-- **MCP Documentation**: https://code.claude.com/docs/mcp
+- **GitHub**: https://github.com/kevintseng/toonify-mcp
+- **Issues**: https://github.com/kevintseng/toonify-mcp/issues
+- **MCP Docs**: https://code.claude.com/docs/mcp
 
----
+## License
 
-**⭐ If this tool saves you money on API costs, consider starring the repo!**
+MIT License - see [LICENSE](LICENSE)

package/README.zh-TW.md

@@ -1,90 +1,60 @@
-# 🎯 Claude Code Toonify
+# 🎯 Toonify MCP
 
 **[English](README.md) | [繁體中文](README.zh-TW.md)**
 
-**使用 TOON 格式優化，降低 Claude API Token 使用量達 60% 以上**
+MCP 伺服器提供 Token 優化工具，將結構化資料轉換為 TOON 格式。
+平均降低 Claude API Token 使用量達 60% 以上。
 
-一個 MCP (Model Context Protocol) 伺服器，提供**按需**的 Token 優化工具，將結構化資料轉換為 TOON (Token-Oriented Object Notation) 格式。支援任何 MCP 相容的 LLM 客戶端（Claude Code、ChatGPT 等）。
+## 功能特色
 
-⚠️ **重要**：此 MCP 伺服器提供必須明確呼叫的**工具** - 不會自動攔截內容。詳見[使用方式](#-使用方式)。
+- **60%+ Token 削減** - 支援 JSON、CSV、YAML 資料
+- **MCP 整合** - 適用於 Claude Code、Claude Desktop
+- **內建指標** - 本地追蹤 Token 節省量
+- **靜默降級** - 絕不中斷工作流程
 
-## 🌟 功能特色
-
-- **🎯 60%+ Token 削減**：結構化資料平均減少 63.9%
-- **💰 大幅節省成本**：每百萬次 API 呼叫節省約 $1,380
-- **⚡ 極速處理**：典型負載僅 <10ms 額外開銷
-- **🔄 靜默降級**：絕不中斷您的工作流程
-- **📊 內建指標**：本地追蹤節省數據
-- **🔌 MCP 整合**：與 Claude Code 無縫協作
-- **🌍 通用相容**：支援任何 LLM（GPT、Claude、Gemini 等）
-
-## 📊 效能表現
-
-| 格式 | 優化前 | 優化後 | 節省 |
-|------|--------|--------|------|
-| JSON | 247 bytes | 98 bytes | **60%** |
-| CSV  | 180 bytes | 65 bytes | **64%** |
-| YAML | 215 bytes | 89 bytes | **59%** |
-
-**Token 削減範例：**
-```
-JSON (142 tokens)：
-{
-  "products": [
-    {"id": 101, "name": "Laptop Pro", "price": 1299},
-    {"id": 102, "name": "Magic Mouse", "price": 79}
-  ]
-}
-
-TOON (57 tokens，削減 60%)：
-products[2]{id,name,price}:
-  101,Laptop Pro,1299
-  102,Magic Mouse,79
-```
-
-## 🚀 安裝步驟
+## 安裝步驟
 
 ### 1. 安裝套件
 
 ```bash
-npm install -g @ktseng/toonify-mcp
+npm install -g toonify-mcp
 ```
 
 ### 2. 註冊至 Claude Code
 
 ```bash
-# 註冊 MCP 伺服器（user scope - 所有專案可用）
+# User scope（所有專案可用）
 claude mcp add --scope user --transport stdio toonify -- /opt/homebrew/bin/toonify-mcp
 
-# 專案專用註冊
+# 或 Project scope
 claude mcp add --scope project --transport stdio toonify -- /opt/homebrew/bin/toonify-mcp
 ```
 
 ### 3. 驗證安裝
 
 ```bash
-# 檢查 MCP 伺服器是否已註冊並連線
 claude mcp list
-
 # 應顯示：toonify: /opt/homebrew/bin/toonify-mcp - ✓ Connected
 ```
 
-## 📖 使用方式
+## 使用方式
 
-### Claude Code 使用者
+### 選項 A：MCP 工具（手動）
 
-#### 選項 A：MCP 伺服器（手動）
-- ❌ **非自動** - Claude 必須主動呼叫工具
-- ✅ **提示時有效** - "使用 toonify 優化這個資料"
-- ✅ **通用相容性** - 適用於任何 MCP 客戶端
+需要時明確呼叫工具：
 
-#### 選項 B：Claude Code Hook（自動）⭐ 推薦
-- ✅ **完全自動** - 透明攔截工具結果
-- ✅ **零開銷** - 無需手動呼叫
-- ✅ **無縫整合** - 適用於 Read、Grep 等檔案工具
-- ⚠️ **僅限 Claude Code** - 不適用於其他 MCP 客戶端
+```bash
+# 優化內容
+claude mcp call toonify optimize_content '{"content": "..."}'
+
+# 查看統計
+claude mcp call toonify get_stats '{}'
+```
+
+### 選項 B：Claude Code Hook（自動）⭐ 推薦
+
+Claude Code 使用者適用的自動攔截：
 
-**安裝方式**：
 ```bash
 cd hooks/
 npm install
@@ -95,24 +65,9 @@ npm run install-hook
 claude hooks list  # 應顯示：PostToolUse
 ```
 
-詳細設定請參閱 `hooks/README.md`。
+詳見 `hooks/README.md`。
 
-### 手動使用 MCP 工具
-
-```bash
-# 優化內容
-claude mcp call toonify optimize_content '{
-  "content": "{\"products\": [{\"id\": 1, \"name\": \"Test\"}]}",
-  "toolName": "Read"
-}'
-
-# 查看統計資料
-claude mcp call toonify get_stats '{}'
-```
-
-## 🔧 設定選項
-
-透過環境變數或設定檔進行配置：
+## 設定選項
 
 ```bash
 # 環境變數
@@ -121,7 +76,7 @@ export TOONIFY_MIN_TOKENS=50
 export TOONIFY_MIN_SAVINGS=30
 export TOONIFY_SKIP_TOOLS="Bash,Write"
 
-# 或使用 ~/.claude/toonify-config.json
+# 或 ~/.claude/toonify-config.json
 {
   "enabled": true,
   "minTokensThreshold": 50,
@@ -130,195 +85,12 @@ export TOONIFY_SKIP_TOOLS="Bash,Write"
 }
 ```
 
-## 📊 統計儀表板
-
-查看您的節省成果：
-
-```bash
-claude mcp call toonify get_stats '{}'
-```
-
-輸出範例：
-```
-📊 Token Optimization Stats
-━━━━━━━━━━━━━━━━━━━━━━━━
-Total Requests: 1,234
-Optimized: 856 (69.4%)
-
-Tokens Before: 1,456,789
-Tokens After: 567,234
-Total Savings: 889,555 (61.1%)
-
-💰 Cost Savings (at $3/1M input tokens):
-   $2.67 saved
-```
-
-## 🏗️ 架構圖
-
-```
-┌──────────────────┐
-│   Claude Code    │
-│   CLI            │
-└─────────┬────────┘
-          │
-          ↓
-┌──────────────────┐
-│  Toonify MCP     │
-│  Server          │
-├──────────────────┤
-│ • optimize_content│ ← 壓縮結構化資料
-│ • get_stats      │ ← 查看指標
-└─────────┬────────┘
-          │
-          ↓
-┌──────────────────┐
-│ TokenOptimizer   │ ← TOON 編碼
-└──────────────────┘
-          │
-          ↓
-┌──────────────────┐
-│ MetricsCollector │ ← 追蹤節省量
-└──────────────────┘
-```
-
-## 🌍 相容性
-
-### ✅ **此 MCP 伺服器支援：**
-- **Claude Code CLI**（主要目標）
-- **Claude Desktop App**
-- **自訂 MCP 客戶端**
-- **任何實作 MCP 協定的工具**
-
-**重要**：MCP（Model Context Protocol）是 Anthropic 的協定。此 MCP 伺服器僅適用於 Claude 生態系統中的 MCP 相容客戶端。
-
-### 🔧 **在其他 LLM 使用 TOON 格式**
-
-雖然此 **MCP 伺服器**僅限 Claude 使用，但 **TOON 格式本身**可為任何 LLM（GPT、Gemini、Llama 等）減少 Token 使用量。若要在非 MCP LLM 使用 TOON 優化：
-
-**TypeScript/JavaScript：**
-```typescript
-import { encode, decode } from '@toon-format/toon';
-
-// 在傳送至任何 LLM API 前優化資料
-const data = {
-  products: [
-    { id: 101, name: 'Laptop Pro', price: 1299 },
-    { id: 102, name: 'Magic Mouse', price: 79 }
-  ]
-};
-
-const optimizedContent = encode(data); // 減少 60% tokens
-
-// 用於 OpenAI
-await openai.chat.completions.create({
-  model: 'gpt-4',
-  messages: [{ role: 'user', content: `分析：${optimizedContent}` }]
-});
-
-// 用於 Gemini
-await gemini.generateContent({
-  contents: [{ text: `分析：${optimizedContent}` }]
-});
-```
-
-**Python：**
-```python
-# 安裝：pip install toonify
-from toonify import encode
-import openai
-
-data = {"products": [...]}
-optimized = encode(data)
-
-# 適用於任何 LLM API
-openai.chat.completions.create(
-    model="gpt-4",
-    messages=[{"role": "user", "content": f"分析：{optimized}"}]
-)
-```
-
-### 📊 **MCP 伺服器 vs TOON 函式庫**
-
-| 功能 | 此 MCP 伺服器 | 直接使用 TOON 函式庫 |
-|------|-------------|-------------------|
-| **目標** | Claude Code/Desktop | 任何 LLM |
-| **整合方式** | 自動（透過 MCP） | 手動（程式碼整合） |
-| **設定** | 配置一次 | 每個專案都要導入 |
-| **相容性** | 僅限 MCP 客戶端 | 通用 |
-
-## 🧪 開發指南
-
-```bash
-# 複製儲存庫
-git clone https://github.com/kevintseng/toonify-mcp.git
-cd toonify-mcp
-
-# 安裝依賴
-npm install
-
-# 建置
-npm run build
-
-# 執行測試
-npm test
-
-# 開發模式（監視變更）
-npm run dev
-```
-
-## 🎯 適用時機
-
-**✅ 最適合用於：**
-- 工具回傳的大型 JSON 回應
-- CSV/表格資料
-- 結構化 API 回應
-- 資料庫查詢結果
-- 包含結構化資料的檔案內容
-
-**⚠️ 不建議用於：**
-- 純文字內容
-- 程式碼檔案
-- 高度不規則的資料結構
-- 內容 < 50 tokens
-
-## 📚 技術細節
-
-### 運作原理
-
-1. **攔截**：MCP 伺服器透過 `optimize_content` 工具攔截工具結果
-2. **偵測**：分析內容以識別 JSON、CSV 或 YAML
-3. **優化**：使用 `@toon-format/toon` 轉換為 TOON 格式
-4. **驗證**：確保節省量 > 30% 閾值
-5. **降級**：若優化失敗或節省量過低則返回原始內容
-
-### 依賴套件
-
-- `@modelcontextprotocol/sdk` - MCP 伺服器框架
-- `@toon-format/toon` - TOON 格式編碼/解碼
-- `tiktoken` - Token 計數（相容 Claude/GPT）
-- `yaml` - YAML 解析支援
-
-## 🤝 貢獻
-
-歡迎貢獻！請參閱 [CONTRIBUTING.md](CONTRIBUTING.md) 了解指南。
-
-## 📄 授權
-
-MIT License - 詳見 [LICENSE](LICENSE) 檔案
-
-## 🙏 致謝
-
-- [Toonify](https://github.com/ScrapeGraphAI/toonify) by ScrapeGraphAI 團隊
-- [Claude Code](https://github.com/anthropics/claude-code) by Anthropic
-- 原始靈感來源：[awesome-llm-apps](https://github.com/Shubhamsaboo/awesome-llm-apps)
-
-## 🔗 連結
+## 連結
 
-- **NPM 套件**：即將推出
 - **GitHub**：https://github.com/kevintseng/toonify-mcp
 - **問題回報**：https://github.com/kevintseng/toonify-mcp/issues
 - **MCP 文件**：https://code.claude.com/docs/mcp
 
----
+## 授權
 
-**⭐ 如果這個工具幫您節省了 API 成本，歡迎給個星星！**
+MIT License - 詳見 [LICENSE](LICENSE)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "toonify-mcp",
-  "version": "0.1.0",
+  "version": "0.1.1",
   "description": "MCP server for TOON format token optimization - works with any MCP client (Claude Code, ChatGPT, etc.)",
   "type": "module",
   "main": "dist/index.js",

package/jest.config.js

@@ -1,23 +0,0 @@
-/** @type {import('jest').Config} */
-export default {
-  preset: 'ts-jest/presets/default-esm',
-  testEnvironment: 'node',
-  extensionsToTreatAsEsm: ['.ts'],
-  moduleNameMapper: {
-    '^(\\.{1,2}/.*)\\.js$': '$1',
-  },
-  transform: {
-    '^.+\\.tsx?$': [
-      'ts-jest',
-      {
-        useESM: true,
-      },
-    ],
-  },
-  transformIgnorePatterns: [
-    'node_modules/(?!(@toon-format|tiktoken)/)',
-  ],
-  testMatch: ['**/tests/**/*.test.ts'],
-  collectCoverageFrom: ['src/**/*.ts'],
-  coveragePathIgnorePatterns: ['/node_modules/', '/dist/'],
-};

package/src/index.ts

@@ -1,23 +0,0 @@
-#!/usr/bin/env node
-
-/**
- * Claude Code Toonify MCP Server
- *
- * Optimizes token usage by converting structured data to TOON format
- * before sending to Claude API, achieving 60%+ token reduction.
- */
-
-import { ToonifyMCPServer } from './server/mcp-server.js';
-
-async function main() {
-  const server = new ToonifyMCPServer();
-
-  try {
-    await server.start();
-  } catch (error) {
-    console.error('Failed to start Toonify MCP server:', error);
-    process.exit(1);
-  }
-}
-
-main();

package/src/metrics/metrics-collector.ts

@@ -1,125 +0,0 @@
-/**
- * MetricsCollector: Local-only metrics tracking
- */
-
-import { promises as fs } from 'fs';
-import path from 'path';
-import os from 'os';
-import type { TokenStats } from '../optimizer/types.js';
-
-export interface OptimizationMetric {
-  timestamp: string;
-  toolName: string;
-  originalTokens: number;
-  optimizedTokens: number;
-  savings: number;
-  savingsPercentage: number;
-  wasOptimized: boolean;
-  format?: string;
-  reason?: string;
-}
-
-export class MetricsCollector {
-  private statsPath: string;
-
-  constructor() {
-    // Store in user's home directory, not in project
-    this.statsPath = path.join(
-      os.homedir(),
-      '.claude',
-      'token_stats.json'
-    );
-  }
-
-  /**
-   * Record an optimization attempt
-   */
-  async record(metric: OptimizationMetric): Promise<void> {
-    try {
-      const stats = await this.loadStats();
-
-      // Update aggregated stats
-      stats.totalRequests++;
-      if (metric.wasOptimized) {
-        stats.optimizedRequests++;
-      }
-
-      stats.tokensBeforeOptimization += metric.originalTokens;
-      stats.tokensAfterOptimization += metric.optimizedTokens;
-      stats.totalSavings += metric.savings;
-
-      // Recalculate average
-      stats.averageSavingsPercentage =
-        stats.optimizedRequests > 0
-          ? (stats.totalSavings / stats.tokensBeforeOptimization) * 100
-          : 0;
-
-      await this.saveStats(stats);
-    } catch (error) {
-      // Silent failure - metrics should never break functionality
-      console.error('Failed to record metrics:', error);
-    }
-  }
-
-  /**
-   * Get current statistics
-   */
-  async getStats(): Promise<TokenStats> {
-    return await this.loadStats();
-  }
-
-  /**
-   * Load stats from disk
-   */
-  private async loadStats(): Promise<TokenStats> {
-    try {
-      // Ensure directory exists
-      await fs.mkdir(path.dirname(this.statsPath), { recursive: true });
-
-      const data = await fs.readFile(this.statsPath, 'utf-8');
-      return JSON.parse(data);
-    } catch {
-      // Return empty stats if file doesn't exist
-      return {
-        totalRequests: 0,
-        optimizedRequests: 0,
-        tokensBeforeOptimization: 0,
-        tokensAfterOptimization: 0,
-        totalSavings: 0,
-        averageSavingsPercentage: 0,
-      };
-    }
-  }
-
-  /**
-   * Save stats to disk
-   */
-  private async saveStats(stats: TokenStats): Promise<void> {
-    await fs.writeFile(
-      this.statsPath,
-      JSON.stringify(stats, null, 2),
-      'utf-8'
-    );
-  }
-
-  /**
-   * Format stats as dashboard
-   */
-  async formatDashboard(): Promise<string> {
-    const stats = await this.getStats();
-
-    return `
-📊 Token Optimization Stats
-━━━━━━━━━━━━━━━━━━━━━━━━
-Total Requests: ${stats.totalRequests}
-Optimized: ${stats.optimizedRequests} (${((stats.optimizedRequests / stats.totalRequests) * 100).toFixed(1)}%)
-
-Tokens Before: ${stats.tokensBeforeOptimization.toLocaleString()}
-Tokens After: ${stats.tokensAfterOptimization.toLocaleString()}
-Total Savings: ${stats.totalSavings.toLocaleString()} (${stats.averageSavingsPercentage.toFixed(1)}%)
-
-💰 Cost Savings (at $3/1M input tokens):
-   $${((stats.totalSavings / 1_000_000) * 3).toFixed(2)} saved
-`.trim();
-  }
-}

package/src/optimizer/token-optimizer.ts

@@ -1,214 +0,0 @@
-/**
- * TokenOptimizer: Core optimization logic using Toonify
- */
-
-import { encode as toonEncode, decode as toonDecode } from '@toon-format/toon';
-import { encoding_for_model } from 'tiktoken';
-import yaml from 'yaml';
-import type {
-  OptimizationResult,
-  ToolMetadata,
-  StructuredData,
-  OptimizationConfig
-} from './types.js';
-
-export class TokenOptimizer {
-  private config: OptimizationConfig;
-  private tokenEncoder;
-
-  constructor(config: Partial<OptimizationConfig> = {}) {
-    this.config = {
-      enabled: true,
-      minTokensThreshold: 50,
-      minSavingsThreshold: 30,
-      maxProcessingTime: 50,
-      skipToolPatterns: [],
-      ...config
-    };
-
-    // Use Claude tokenizer
-    this.tokenEncoder = encoding_for_model('gpt-4');
-  }
-
-  /**
-   * Main optimization method
-   */
-  async optimize(
-    content: string,
-    metadata?: ToolMetadata
-  ): Promise<OptimizationResult> {
-    const startTime = Date.now();
-
-    // Quick path: skip if disabled or content too small
-    if (!this.config.enabled || content.length < 200) {
-      return {
-        optimized: false,
-        originalContent: content,
-        originalTokens: this.countTokens(content),
-        reason: 'Content too small'
-      };
-    }
-
-    // Skip if tool matches skip patterns
-    if (metadata?.toolName && this.shouldSkipTool(metadata.toolName)) {
-      return {
-        optimized: false,
-        originalContent: content,
-        originalTokens: this.countTokens(content),
-        reason: `Tool ${metadata.toolName} in skip list`
-      };
-    }
-
-    // Detect structured data
-    const structuredData = this.detectStructuredData(content);
-    if (!structuredData) {
-      return {
-        optimized: false,
-        originalContent: content,
-        originalTokens: this.countTokens(content),
-        reason: 'Not structured data'
-      };
-    }
-
-    try {
-      // Convert to TOON format
-      const toonContent = toonEncode(structuredData.data);
-
-      // Count tokens
-      const originalTokens = this.countTokens(content);
-      const optimizedTokens = this.countTokens(toonContent);
-
-      // Calculate savings
-      const tokenSavings = originalTokens - optimizedTokens;
-      const savingsPercentage = (tokenSavings / originalTokens) * 100;
-
-      // Check if worth using
-      if (savingsPercentage < this.config.minSavingsThreshold) {
-        return {
-          optimized: false,
-          originalContent: content,
-          originalTokens,
-          reason: `Savings too low: ${savingsPercentage.toFixed(1)}%`
-        };
-      }
-
-      // Check processing time
-      const elapsed = Date.now() - startTime;
-      if (elapsed > this.config.maxProcessingTime) {
-        return {
-          optimized: false,
-          originalContent: content,
-          originalTokens,
-          reason: `Processing timeout: ${elapsed}ms`
-        };
-      }
-
-      return {
-        optimized: true,
-        originalContent: content,
-        optimizedContent: toonContent,
-        originalTokens,
-        optimizedTokens,
-        savings: {
-          tokens: tokenSavings,
-          percentage: savingsPercentage
-        },
-        format: structuredData.type
-      };
-
-    } catch (error) {
-      // Silent fallback on error
-      return {
-        optimized: false,
-        originalContent: content,
-        originalTokens: this.countTokens(content),
-        reason: `Error: ${error instanceof Error ? error.message : 'Unknown'}`
-      };
-    }
-  }
-
-  /**
-   * Detect if content is structured data (JSON/CSV/YAML)
-   */
-  private detectStructuredData(content: string): StructuredData | null {
-    // Try JSON first
-    try {
-      const data = JSON.parse(content);
-      if (typeof data === 'object' && data !== null) {
-        return { type: 'json', data, confidence: 1.0 };
-      }
-    } catch {}
-
-    // Try YAML
-    try {
-      const data = yaml.parse(content);
-      if (typeof data === 'object' && data !== null) {
-        return { type: 'yaml', data, confidence: 0.9 };
-      }
-    } catch {}
-
-    // Try CSV (simple heuristic)
-    if (this.looksLikeCSV(content)) {
-      try {
-        const data = this.parseSimpleCSV(content);
-        return { type: 'csv', data, confidence: 0.8 };
-      } catch {}
-    }
-
-    return null;
-  }
-
-  /**
-   * Simple CSV detection heuristic
-   */
-  private looksLikeCSV(content: string): boolean {
-    const lines = content.split('\n').filter(l => l.trim());
-    if (lines.length < 2) return false;
-
-    const firstLineCommas = (lines[0].match(/,/g) || []).length;
-    if (firstLineCommas === 0) return false;
-
-    // Check if most lines have similar comma count
-    let matchingLines = 0;
-    for (let i = 1; i < Math.min(lines.length, 10); i++) {
-      const commas = (lines[i].match(/,/g) || []).length;
-      if (commas === firstLineCommas) matchingLines++;
-    }
-
-    return matchingLines >= Math.min(lines.length - 1, 7);
-  }
-
-  /**
-   * Parse simple CSV to array of objects
-   */
-  private parseSimpleCSV(content: string): any[] {
-    const lines = content.split('\n').filter(l => l.trim());
-    const headers = lines[0].split(',').map(h => h.trim());
-
-    return lines.slice(1).map(line => {
-      const values = line.split(',').map(v => v.trim());
-      const obj: any = {};
-      headers.forEach((header, i) => {
-        obj[header] = values[i] || '';
-      });
-      return obj;
-    });
-  }
-
-  /**
-   * Count tokens in text
-   */
-  private countTokens(text: string): number {
-    return this.tokenEncoder.encode(text).length;
-  }
-
-  /**
-   * Check if tool should be skipped
-   */
-  private shouldSkipTool(toolName: string): boolean {
-    return this.config.skipToolPatterns?.some(pattern => {
-      const regex = new RegExp(pattern);
-      return regex.test(toolName);
-    }) ?? false;
-  }
-}

package/src/optimizer/types.ts

@@ -1,46 +0,0 @@
-/**
- * Type definitions for token optimization
- */
-
-export interface OptimizationResult {
-  optimized: boolean;
-  originalContent: string;
-  optimizedContent?: string;
-  originalTokens: number;
-  optimizedTokens?: number;
-  savings?: {
-    tokens: number;
-    percentage: number;
-  };
-  format?: 'json' | 'csv' | 'yaml' | 'unknown';
-  reason?: string; // Why optimization was skipped
-}
-
-export interface ToolMetadata {
-  toolName: string;
-  contentType?: string;
-  size: number;
-}
-
-export interface StructuredData {
-  type: 'json' | 'csv' | 'yaml';
-  data: any;
-  confidence: number;
-}
-
-export interface OptimizationConfig {
-  enabled: boolean;
-  minTokensThreshold: number; // Only optimize if content > N tokens
-  minSavingsThreshold: number; // Only use if savings > N%
-  maxProcessingTime: number; // Max ms to spend optimizing
-  skipToolPatterns?: string[]; // Tool names to skip
-}
-
-export interface TokenStats {
-  totalRequests: number;
-  optimizedRequests: number;
-  tokensBeforeOptimization: number;
-  tokensAfterOptimization: number;
-  totalSavings: number;
-  averageSavingsPercentage: number;
-}

package/src/server/mcp-server.ts

@@ -1,134 +0,0 @@
-/**
- * ToonifyMCPServer: MCP server that wraps tool results with token optimization
- */
-
-import { Server } from '@modelcontextprotocol/sdk/server/index.js';
-import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
-import {
-  CallToolRequestSchema,
-  ListToolsRequestSchema,
-} from '@modelcontextprotocol/sdk/types.js';
-import { TokenOptimizer } from '../optimizer/token-optimizer.js';
-import { MetricsCollector } from '../metrics/metrics-collector.js';
-
-export class ToonifyMCPServer {
-  private server: Server;
-  private optimizer: TokenOptimizer;
-  private metrics: MetricsCollector;
-
-  constructor() {
-    this.server = new Server(
-      {
-        name: 'claude-code-toonify',
-        version: '0.1.0',
-      },
-      {
-        capabilities: {
-          tools: {},
-        },
-      }
-    );
-
-    this.optimizer = new TokenOptimizer();
-    this.metrics = new MetricsCollector();
-
-    this.setupHandlers();
-  }
-
-  private setupHandlers() {
-    // List available tools
-    this.server.setRequestHandler(ListToolsRequestSchema, async () => ({
-      tools: [
-        {
-          name: 'optimize_content',
-          description: 'Optimize structured data content for token efficiency using TOON format',
-          inputSchema: {
-            type: 'object',
-            properties: {
-              content: {
-                type: 'string',
-                description: 'The content to optimize (JSON, CSV, or YAML)',
-              },
-              toolName: {
-                type: 'string',
-                description: 'Name of the tool that generated this content',
-              },
-            },
-            required: ['content'],
-          },
-        },
-        {
-          name: 'get_stats',
-          description: 'Get token optimization statistics',
-          inputSchema: {
-            type: 'object',
-            properties: {},
-          },
-        },
-      ],
-    }));
-
-    // Handle tool calls
-    this.server.setRequestHandler(CallToolRequestSchema, async (request) => {
-      const { name, arguments: args } = request.params;
-
-      switch (name) {
-        case 'optimize_content': {
-          const { content, toolName } = args as {
-            content: string;
-            toolName?: string;
-          };
-
-          const result = await this.optimizer.optimize(content, {
-            toolName: toolName || 'unknown',
-            size: content.length,
-          });
-
-          // Record metrics
-          await this.metrics.record({
-            timestamp: new Date().toISOString(),
-            toolName: toolName || 'unknown',
-            originalTokens: result.originalTokens,
-            optimizedTokens: result.optimizedTokens || result.originalTokens,
-            savings: result.savings?.tokens || 0,
-            savingsPercentage: result.savings?.percentage || 0,
-            wasOptimized: result.optimized,
-            format: result.format,
-            reason: result.reason,
-          });
-
-          return {
-            content: [
-              {
-                type: 'text',
-                text: JSON.stringify(result, null, 2),
-              },
-            ],
-          };
-        }
-
-        case 'get_stats': {
-          const stats = await this.metrics.getStats();
-          return {
-            content: [
-              {
-                type: 'text',
-                text: JSON.stringify(stats, null, 2),
-              },
-            ],
-          };
-        }
-
-        default:
-          throw new Error(`Unknown tool: ${name}`);
-      }
-    });
-  }
-
-  async start() {
-    const transport = new StdioServerTransport();
-    await this.server.connect(transport);
-
-    console.error('Toonify MCP Server running on stdio');
-  }
-}

package/test-mcp.json

@@ -1,5 +0,0 @@
-{
-  "jsonrpc": "2.0",
-  "id": 1,
-  "method": "tools/list"
-}