Package Exports

@sid7vish/universal-rag-mcp
@sid7vish/universal-rag-mcp/dist/index.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@sid7vish/universal-rag-mcp) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

Universal RAG MCP

Intelligent cross-platform memory system for AI assistants via Model Context Protocol

Give your AI assistants a persistent, searchable memory that works across Claude, ChatGPT, Gemini, and more. Features intelligent chunking, automatic deduplication, and semantic search.

✨ Features

🧠 Smart Memory - Intelligent chunking filters noise, keeps only important information
🔍 Semantic Search - Find information by meaning, not just keywords
🚫 Auto-Deduplication - Tracks mention count instead of saving duplicates
🌐 Cross-Platform - Same memory across Claude, ChatGPT, Gemini, etc.
⚡ Fast - 2-second indexing, sub-100ms searches
🎯 Accurate - Multi-question support with parallel searches
🔒 Your Data - Stored in your Firebase/Pinecone accounts

Features

Cross-platform: Same memory in Claude Desktop, ChatGPT, Gemini, and more
Your data: You control it - stored in your Firebase/Pinecone accounts
Zero config: 5-minute setup wizard handles everything
Smart search: Semantic search with automatic reranking
Fast: In-memory cache + hot/warm/cold storage tiers

Quick Start

1. Install

npm install -g universal-rag-mcp

2. Setup (5 minutes)

universal-rag-mcp init

You'll need 3 free API keys:

Firebase (database) - https://console.firebase.google.com
Pinecone (vector search) - https://app.pinecone.io/signup
Voyage AI (embeddings) - https://dash.voyageai.com/api-keys

3. Use

Restart Claude Desktop (or your AI platform), then:

You: Remember that I love TypeScript
AI: I'll remember that!

You: What programming languages do I like?
AI: Based on your memory, you love TypeScript!

How It Works

Add to memory: Conversations are automatically chunked and embedded
Search memory: Semantic search finds relevant context from past conversations
Storage tiers:
- Hot: Last 10 queries (5ms access)
- Warm: Recent conversations (50-200ms)
- Cold: Old conversations (200-500ms)

Architecture

┌─────────────────┐
│  AI Platform    │ (Claude, ChatGPT, etc.)
└────────┬────────┘
         │ MCP Protocol
┌────────▼────────┐
│  This Tool      │
└────────┬────────┘
         │
    ┌────┴────┬────────┬──────────┐
    ▼         ▼        ▼          ▼
Firebase  Pinecone  Voyage   In-Memory
(data)    (vectors) (embed)   (cache)

API Keys Needed

Mandatory (3)

Firebase Firestore - Store conversation data
- Free tier: 1GB storage, 50K reads/day
- Setup: 2 minutes
Pinecone - Vector search
- Free tier: 5M vectors
- Setup: 1 minute
Voyage AI - Generate embeddings
- Free tier: 10M tokens
- Setup: 1 minute

Optional (0)

None! We use built-in MMR reranking and in-memory caching instead of Cohere/Redis.

Commands

# Initial setup
universal-rag-mcp init

# Check status
universal-rag-mcp status

# Show config location
universal-rag-mcp config

MCP Tools

search_memory

Search your conversation history:

{
  "query": "What do I like?",
  "limit": 5,
  "minScore": 0.6,
  "platform": "claude"
}

add_to_memory

Store new content:

{
  "content": "I love TypeScript",
  "platform": "claude",
  "role": "user"
}

Configuration

Config stored at: ~/.config/universal-rag-mcp/config.json

{
  "userId": "your-email@example.com",
  "firebase": {
    "projectId": "your-project",
    "serviceAccountPath": "/path/to/key.json"
  },
  "pinecone": {
    "apiKey": "your-key",
    "indexName": "rag-your-id"
  },
  "voyage": {
    "apiKey": "your-key"
  }
}

Development

# Clone repo
git clone https://github.com/yourusername/universal-rag-mcp
cd universal-rag-mcp

# Install dependencies
npm install

# Build
npm run build

# Run
npm start

License

MIT