Building an “Agentic Middleware” for the City Government: Boston’s Experiment with Model Context Protocol
Boston is preparing for a future where AI agents increasingly interact with government systems. In this interview, Boston's Chief Information Officer Santiago Garces explains how the city is experimenting with the Model Context Protocol (MCP) as a governance layer between AI and public digital infrastructure. Starting with the open data portal, Boston is testing how MCP can make AI interactions more reliable, secure, and grounded in real government data. The effort provides a concrete example of how to safely support the emerging “agentic web” while improving access to public services.