Revolutionizing Dataverse Data Management with Model Context Protocol (MCP) for AI

Abstract

At the forefront of evolving AI infrastructure, we’re building a new agentic architecture for Dataverse that bridges structured data, language models, and distributed knowledge systems. From multilingual metadata harvesting to massive interactive knowledge graphs, our journey began with the Dutch national project ODISSEI, which laid the groundwork for AI-ready, multilingual data discovery. We now support Croissant (soon a standard at NeurIPS) — a semantic layer enabling LLMs to reason over datasets with JSON-LD. Tools like DataChat and Ask Dataverse already show what’s possible with structured prompts and reasoning over live metadata.