Otzaria Server

An MCP (Model Context Protocol) server that provides powerful search capabilities for Jewish texts and literature. This server enables Large Language Models to search and reference Jewish texts through a standardized interface.

Features

• Full-text search across Jewish texts and literature
• Advanced query syntax support:
  • Field-specific search (text:term, reference:term, topics:term)
  • Boolean operators (AND, OR)
  • Required/excluded terms (+term, -term)
  • Phrase search ('exact phrase')
  • Wildcards (?, *)
• Relevance-based scoring
• Rich search results including references, topics, and highlighted excerpts

Installation

Requires Python 3.10 or higher.

Clone the repository

git clone https://github.com/sivan22/mcp-otzaria-server.git
cd mcp-otzaria-server

Get the index

download and extract the index from here

Install dependencies

pip install .

Running the Server

The server can be run directly:

uv --directory path/to/directory run jewish_library

Or through an MCP client that supports the Model Context Protocol. for claude desktop app and cline you should use the following config:

{
  "mcpServers": {        
      "jewish_library": {
          "command": "uv",
          "args": [
              "--directory",
              "your/path/to/directory",
              "run",
              "jewish_library"
          ],
          "env": {
            "PYTHONIOENCODING": "utf-8" 
          }
      }
  }
}

Available tools

The server provides a single tool through the MCP interface:

full_text_search

Performs a full-text search across the Jewish library with advanced query capabilities.

Example query formats:

# Basic search
"maimonides on prayer"

# Field-specific search
text:"love your neighbor" AND topics:mitzvot

# Required terms
+shabbat +candles

# Phrase search with topic filter
"four species" AND topics:sukkot

# Wildcard search
pray* AND reference:psalms

Search results include:

• Reference information
• Relevant topics
• Highlighted excerpts showing query matches
• Relevance score

Development

This project uses:

• MCP SDK for server implementation
• Tantivy for full-text search capabilities

Requirements

• Python >= 3.10
• MCP SDK >= 1.1.1
• Tantivy search engine

License

MIT License