mcp-local-rag

Overview

mcp-local-rag is a privacy-focused RAG solution that runs entirely on the user's machine. It offers semantic search with keyword boost to catch exact technical terms like function names or error codes that might be missed by pure vector search. The server provides 7 tools including document ingestion, querying, context expansion, and file management. It supports various document formats including PDF, DOCX, TXT, and Markdown, with an option to ingest HTML content via its API.

Try asking AI

After installing, here are 7 things you can ask your AI assistant:

• ingest_file

  Ingest a document file into the RAG database

  Try:Ingest the document at /Users/me/docs/api-spec.pdfTry:Add the project documentation to the RAG system
• ingest_data

  Ingest HTML content into the RAG database

  Try:Fetch https://example.com/docs and ingest the HTMLTry:Process this HTML content I retrieved
• query_documents

  Search documents using semantic similarity with keyword boost

  Try:What does the API documentation say about authentication?Try:Find information about rate limitingTry:Search for error handling best practices
• read_chunk_neighbors

  Read chunks surrounding a specific search result

  Try:That result about authentication looks relevant — read the surrounding chunksTry:Get more context around this result
• list_files

  List all files in BASE_DIR and their ingested status

  Try:List all files in my documents directoryTry:Show which files have been ingested
• delete_file

  Delete a file from the RAG database

  Try:Delete old-spec.pdf from RAGTry:Remove this document from the index
• status

  Check RAG server system health and database stats

  Try:Show RAG server statusTry:Check how many documents are indexed

Comparable tools

Installation

{
  "mcpServers": {
    "local-rag": {
      "command": "npx",
      "args": ["-y", "mcp-local-rag"],
      "env": {
        "BASE_DIR": "/path/to/your/documents"
      }
    }
  }
}

claude mcp add local-rag --scope user --env BASE_DIR=/path/to/your/documents -- npx -y mcp-local-rag

npx mcp-local-rag ingest ./docs/
npx mcp-local-rag query "authentication API"

FAQ

What document formats are supported?

The server supports PDF, DOCX, TXT, and Markdown documents. It can also ingest HTML content via the ingest_data tool.

How does keyword boost work?

Keyword boost is applied after semantic filtering. It re-ranks results to prioritize exact matches of technical terms like function names, class names, or error codes while maintaining semantic relevance.

Is there a web interface?

No, this is a CLI and MCP server-only solution that runs locally without a web interface.

Compare mcp-local-rag with