Agentic AI in Data Commons and GLAM: What Role for the Model Context Protocol?

Abstract

The emergence of the agentic web, in which AI agents autonomously navigate online services to accomplish complex tasks, poses urgent questions for digital commons and GLAM institutions (Galleries, Libraries, Archives, and Museums). This paper reports on a crosssectoral workshop held at the Bibliothèque nationale de France, which examined the Model Context Protocol (MCP) as a new infrastructure layer for regulating agentic access to open knowledge resources. Drawing on presentations by practitioners and researchers across the library, archive, museum, performing arts, open knowledge, and open science sectors, the paper documents emerging MCP experiments, identifies transversal governance challenges, and synthesizes the outcomes of a collective prospective scenario-building exercise. Key findings include: the reframing of MCP server deployment as an editorial act with significant implications for data representation and institutional authority; the identification of critical capacity gaps between API-ready and non-API-ready organizations; and the recognition of MCP governance as a site of broader advocacy for the commons and public goods within the agentic web.