MCP Server for PDF Q&A
Grounding LLM answers in your own documents
A Model Context Protocol server that connects PDF documents to LLMs through an automated retrieval pipeline — extraction, chunking, embedding, and similarity search.
Overview
LLMs answer confidently whether or not they know the answer. This project grounds them in real documents: a Model Context Protocol (MCP) server that exposes a PDF corpus to any MCP-compatible LLM client.
Documents flow through an automated ETL pipeline — PDF and Markdown extraction, overlapping sliding-window chunking, then embedding — and are indexed for similarity-based retrieval, so each answer is assembled from the passages that actually support it.
The conversational interface supports multi-turn Q&A over large document collections, resolving follow-up questions against earlier answers and retrieved context.