Quarry

About

Quarry helps users unlock the knowledge hidden within their local files, whether PDFs, images, text documents, or source code. It builds a private, local knowledge base using a vector database, allowing for powerful semantic search that understands meaning, not just keywords. Designed to integrate seamlessly with LLMs via the Model Context Protocol (MCP), Quarry empowers tools like Claude Desktop and Claude Code to access and leverage your indexed documents for enhanced conversations and code generation, all while keeping your data local and private by default.

Key Features

• Operate fully offline with local OCR and embedding models, offering optional cloud backends for enhanced OCR or large-scale embedding.
• Organize content using named databases, collections, and automatic directory synchronization.
• 0 GitHub stars
• Ingest a wide range of file types: PDFs (text & image), images, spreadsheets, presentations, HTML, text, and 30+ programming languages.
• Perform semantic search, understanding the meaning behind your queries to find relevant passages even without exact keyword matches.
• Serve as an MCP server, providing LLMs like Claude Desktop and Claude Code direct access to your indexed content for contextual understanding.

Use Cases

• Building a personal, searchable knowledge base from diverse local documents like reports, notes, and code.
• Providing large language models (LLMs) with private, real-time context from your hard drive for improved responses.
• Efficiently finding information across multiple file types using semantic search without relying on cloud services.