notwhiteblank/scholar-rag-mcp

Overview

An MCP server that turns academic papers into a searchable knowledge base. It ingests PDFs with MinerU, stores vectors in Qdrant, and applies reranking to search results. It also supports a two-phase knowledge base management workflow.

Best for

Best for
Researchers and developers building paper Q&A or retrieval systems

Use cases

• Build a personal searchable library of research papers
• Retrieve relevant paper passages with reranked vector search
• Manage a two-phase knowledge base for academic document collections

How to use

Install

uvx scholar-rag-mcp             # download and start the MCP server (stdio)

Tools exposed

• list_kbs
• create_kb
• delete_kb
• add_document
• remove_document
• get_document
• get_document_text
• list_documents
• search_documents
• search_chunks
• get_job

Tested with

Claude Desktop, Claude Code, ChatGPT

Notes

An MCP server that turns academic papers into a searchable knowledge base. It ingests PDFs with MinerU, stores vectors in Qdrant, and applies reranking to search results. It also supports a two-phase knowledge base management workflow.

4 stars on GitHub. Last updated 2026-09-04. Licensed MIT.

Use cases

• Build a personal searchable library of research papers
• Retrieve relevant paper passages with reranked vector search
• Manage a two-phase knowledge base for academic document collections

Pros

• Uses a modern RAG stack with reranking for better retrieval
• Open source and Python based, easy to extend
• Designed specifically for academic PDFs

Cons

• Requires separate setup of Qdrant and MinerU
• Python only, limiting integration with non-Python stacks
• Two-phase KB management adds operational complexity

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.