# RagAlgo: Dynamic RAG Engine for AI Reliability [![npm version](https://img.shields.io/npm/v/ragalgo-mcp-server.svg)](https://www.npmjs.com/package/ragalgo-mcp-server) [![npm downloads](https://img.shields.io/npm/dm/ragalgo-mcp-server.svg)](https://www.npmjs.com/package/ragalgo-mcp-server) [![GitHub stars](https://img.shields.io/github/stars/kokogo100/ragalgo-mcp-server?style=social)](https://github.com/kokogo100/ragalgo-mcp-server) [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![MCP](https://img.shields.io/badge/MCP-Compatible-blue)](https://modelcontextprotocol.io) > **"Your AI is an Analyst, NOT a Day Trader."** RagAlgo is an **MCP Server** that provides **mathematically scored financial context** (Korean Stocks/Crypto) to AI agents. We focus on **"State-of-Truth"** (Daily Closed Data) to prevent AI hallucinations caused by real-time market noise. - **Analyst, Not Broker:** We provide "Daily Analysis Reports" (Post-Market), not real-time tick data. - **Scored Context:** Instead of raw prices, we give you "Scores" (0~100) and "Zones" (Forest vs Tree). - **Global Market Specialist:** Optimized for US, UK, JP, KR, and Crypto. πŸ‘‰ **[Official Website (ragalgo.com)](https://www.ragalgo.com)** --- ## πŸ“– Architecture & Whitepaper Discover why RagAlgo is the **"Hippocampus"** for Agentic AI, not just another RAG. ### πŸ—οΈ Data Pipeline Architecture Our production system on Railway processes global financial data 24/7: ``` β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”‚ RagAlgo Data Pipeline (Railway) β”‚ β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€ β”‚ β”‚ β”‚ πŸ“₯ COLLECT πŸ” FILTER 🏷️ TAG πŸ“Š SCORE β”‚ β”‚ ───────────── ────────────── ───────── ────────── β”‚ β”‚ β€’ KR-News-Collector β€’ filter-worker-1 β€’ tag-worker β€’ Gemini-1 β”‚ β”‚ β€’ US-News-Collector β€’ filter-worker-2 β€’ Meta-Hierarchy β€’ Gemini-2 β”‚ β”‚ β€’ UK-News-Collector β€’ filter-worker-3 β€’ Worker β€’ ... β”‚ β”‚ β€’ JP-News-Collector β€’ ibkr_filter_worker3 β”‚ β€’ Gemini-7 β”‚ β”‚ β€’ research-collector β”‚ β”‚ β”‚ β”‚ β”‚ β”‚ β”‚ β”‚ β”‚ β”‚ ════════════════════════════════════════════════════════════════════════ β”‚ β”‚ ↓ β”‚ β”‚ πŸ“¦ SNAPSHOT (Daily 18:00 KST) β”‚ β”‚ ────────────────────────────── β”‚ β”‚ β€’ KR-Snapshot β€’ US-Snapshot β”‚ β”‚ β€’ UK-Snapshot β€’ JP-Snapshot β”‚ β”‚ β€’ CRY-Snapshot β€’ Unified-Snapshot β”‚ β”‚ ↓ β”‚ β”‚ πŸš€ SERVE (MCP Server) β”‚ β”‚ ───────────────────── β”‚ β”‚ β€’ RagAlgo-Service (SSE/stdio) β”‚ β”‚ β€’ ragalgo-relay-server (WebSocket) β”‚ β”‚ β”‚ β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜ ``` ![RagAlgo Architecture](./architecture.png) * **[Vision Whitepaper (The Hook)](./docs/RagAlgo_Report_EN.md)** * **Concept**: Why RagAlgo is a "Semantic Digital Twin" (SDT) using the Hippocampus analogy. * **Value**: Explains the "Self-Growing Taxonomy" and "Data Flywheel" effect. * **[Technical Report (The Proof)](./docs/CKN_Architecture_EN.md)** * **Deep Dive**: Detailed anatomy of the Contextual Knowledge Network (CKN). * **2025 Trend**: How RagAlgo serves as the memory layer for **Agentic AI** (e.g., PepsiCo/Salesforce Agentforce). --- ## πŸ’‘ Why "Daily Close"? Users often ask: *"Why isn't the chart data real-time?"* **Because AI performs better with clarity.** Real-time tick data is full of noise and volatility. If you feed an LLM raw live prices, it often hallucinates patterns that don't exist. RagAlgo acts like a **Professional Technical Analyst** who works after the market closes: 1. **Wait for the dust to settle** (Market Close). 2. **Analyze the day's battle** (Daily Candle & Aux Indicators). 3. **Deliver a "Confirmed Strategy"** to your AI. Use RagAlgo to build **"Investment Advisors"**, not "High-Frequency Trading Bots". --- ## πŸš€ Quick Start ### Claude Desktop Configuration Add this to your config file: - **Windows:** `%APPDATA%\Claude\claude_desktop_config.json` - **Mac:** `~/Library/Application Support/Claude/claude_desktop_config.json` #### ☁️ Cloud Mode (Recommended - No installation required) ```json { "mcpServers": { "ragalgo": { "url": "https://ragalgo-service-production.up.railway.app/sse", "env": { "RAGALGO_API_KEY": "YOUR_API_KEY_HERE" } } } } ``` #### πŸ“¦ Local Mode (Requires Node.js) ```json { "mcpServers": { "ragalgo": { "command": "npx", "args": ["-y", "ragalgo-mcp-server", "--stdio"], "env": { "RAGALGO_API_KEY": "YOUR_API_KEY_HERE" } } } } ``` > **Tip:** You can get a **Free 1,000 Call Key** instantly at [RagAlgo Dashboard](https://www.ragalgo.com/dashboard). --- ## πŸ“š Usage Examples (Cookbook) We have a **dedicated repository** for practical examples to help you get started quickly. Please visit the **[RagAlgo Examples Repository](https://github.com/kokogo100/ragalgo-examples)**. ### What's Inside? - **8 Step-by-Step Recipes:** From basic data fetching to advanced AI agents. - **Skeleton Code + Prompts:** Copy-paste ready resources. - **Scenarios:** - 🐣 **Basic:** Get stock scores in 5 minutes. - πŸ§ͺ **Intermediate:** Verify technical signals with AI. - πŸš€ **Advanced:** Build an autonomous reasoning agent (Mock Trading Audit). - β˜• **Morning Briefing:** Create a bot that emails you a daily market summary. > **"Skeleton + Prompt" Approach:** We provide the ingredients. You ask ChatGPT/Claude to cook! --- ## 🌍 Supported Markets & Roadmap RagAlgo is expanding its CKN coverage globally. Currently, **US, UK, Japan, Korea, and Crypto** markets are fully supported. | Market | Asset Class | Status | | :--- | :--- | :--- | | **πŸ‡°πŸ‡· Korea** | KOSPI / KOSDAQ | **🟒 Live** (Real-time Sentiment & Charts) | | **πŸ‡ΊπŸ‡Έ USA** | NYSE / NASDAQ | **🟒 Live** (Daily Scored Context) | | **πŸ‡―πŸ‡΅ Japan** | Nikkei 225 | **🟒 Live** (Daily Scored Context) | | **πŸ‡¬πŸ‡§ UK** | LSE | **🟒 Live** (Daily Scored Context) | | **πŸͺ™ Crypto** | Global (Upbit/Binance) | **🟒 Live** (Real-time Sentiment & Charts) | --- ## πŸ›  Tools > **⚠️ CORE CONCEPT: Scored vs Raw** > - **`get_news_scored` (Default):** Returns only significant news (Scores β‰  0). Best for AI decision making. > - **`get_news` (Raw):** Returns ALL news including noise. Use this ONLY if you need raw data feed. | Tool | Description | |------|-------------| | `get_news_scored` | **[RECOMMENDED]** News **WITH** AI Sentiment Scores (-10 ~ +10). Filters out noise. | | `get_news` | **[Advanced]** Raw News **WITHOUT** scores. Includes 0-score noise. Use only if you build your own scorer. | | `get_chart_stock` | **[Core]** Global Stock (US/UK/JP/KR) Technical Analysis (Daily Close). | | `get_chart_coin` | **[Core]** Global Crypto Technical Analysis (Daily Close). | | `get_snapshots` | **[Best]** Market Overview (News + Chart + Trend) in one call. | | `get_financials` | Corporate Financials (Quarterly/Yearly). | | `search_tags` | Convert names (e.g., "Samsung") to RagAlgo Tags. | --- ## πŸ“‘ Real-time WebSocket (Business Tier) For users who *really* need live data (e.g., for monitoring dashboards), we offer a WebSocket stream. *Note: This is strictly for monitoring, not for LLM inference context.* - **Access:** Business Plan subscribers (Includes 30 connections). - **Address:** `wss://ragalgo-relay-server-1-production.up.railway.app` - **Guide:** See [Developer Docs](https://www.ragalgo.com/docs) for implementation details. --- ## πŸ’¬ Support - **Website:** [ragalgo.com](https://www.ragalgo.com) - **Email:** support@ragalgo.com