Pageindex Mcp

Pageindex MCP exposes a vectorless, reasoning-based RAG (Retrieval-Augmented Generation) system that represents documents as hierarchical tree structures, enabling LLMs to navigate and retrieve information through logical reasoning rather than vector similarity. The server provides tools for LLMs to interact with local and online PDFs by reasoning over document structure, eliminating the need for vector databases, chunking, or context window limitations when working with long documents. It solves the problem of context overflow and irrelevant retrieval in traditional RAG systems by allowing LLMs to retrieve information like humans navigate a book's index—through structural understanding and multi-step reasoning.