rag-mcp-server by Scarmonit - MCP Server

RAG MCP Server

A Model Context Protocol (MCP) server providing semantic document search capabilities using ChromaDB vector store and ONNX embeddings.

Features

13 MCP Tools for document ingestion, search, and management
Semantic Search using all-MiniLM-L6-v2 embeddings (384 dimensions)
Hybrid Search combining semantic similarity with keyword filtering
ChromaDB Backend with HNSW cosine similarity indexing
Thread-Safe operations with RLock protection
Source Caching for O(1) lookups with 60s TTL

Installation

# Clone the repository
git clone https://github.com/Scarmonit/rag-mcp-server.git
cd rag-mcp-server

# Create virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

MCP Configuration

Add to your MCP client configuration (e.g., Claude Desktop claude_desktop_config.json):

{
  "mcpServers": {
    "rag": {
      "command": "python",
      "args": ["-m", "rag_server.server"],
      "cwd": "/path/to/rag-mcp-server",
      "env": {
        "RAG_DB_PATH": "./rag_chroma_db"
      }
    }
  }
}

Available Tools

Search Tools

Tool	Description
`search_docs`	Semantic document search with optional source and metadata filtering
`hybrid_search`	Combined semantic + keyword search
`search_with_threshold`	Search with minimum similarity score filtering

Ingestion Tools

Tool	Description
`add_document`	Add a single document with metadata
`add_documents_batch`	Batch ingestion with automatic chunking
`ingest_file`	Ingest local file with auto-chunking
`ingest_url`	Fetch and ingest content from URL

Utility Tools

Tool	Description
`chunk_document`	Preview text chunking without ingestion
`list_sources`	List all indexed sources
`delete_source`	Remove all documents from a source
`export_documents`	Export documents to JSON or summary
`get_stats`	Get collection statistics
`health_check`	System health check

Usage Examples

Add a Document

Use the add_document tool with:
- content: "Python is a versatile programming language..."
- source: "python-guide"
- metadata: {"category": "programming", "level": "beginner"}

Search Documents

Use the search_docs tool with:
- query: "What programming languages are good for beginners?"
- n_results: 5
- source_filter: "python-guide" (optional)

Hybrid Search

Use the hybrid_search tool with:
- query: "async programming"
- keywords: ["await", "coroutine"]
- n_results: 10

Environment Variables

Variable	Default	Description
`RAG_DB_PATH`	`./rag_chroma_db`	Path to ChromaDB storage directory

Architecture

rag_server/
├── server.py        # FastMCP server with 13 tools
├── vector_store.py  # ChromaDB wrapper with caching
├── embedding.py     # ONNX embedding service
├── chunking.py      # Text chunking utilities
└── __init__.py      # Package metadata

Testing

# Run all tests
python -m pytest tests/ -v

# Run with coverage
python -m pytest tests/ --cov=rag_server --cov-report=term-missing

License

MIT License - see for details.