The rise of large language models is reshaping how we interact with information, and geographic data should be no exception. In this exploratory talk, we present IGN’s early work on a Geo‑Context MCP, a new interface designed to make France’s sovereign geodata directly accessible to AI agents. The goal is simple but ambitious: allow intelligent systems to query, understand, and reason over geographical data and OGC services—just as easily as humans do today.

We will walk through the first experiments that connect AI agents to WFS endpoints, structured geographic datasets, and other key services from the Géoplateforme. By exposing geodata through a machine‑native contextual layer, IGN aims to lower the barrier between spatial information and AI‑driven analysis. We hope this can open the door to new forms of automated geoprocessing, enriched decision‑making, and dynamic geospatial exploration.

Finally, this talk invites the community to imagine what comes next. IGN’s initiative is intentionally open, experimental, and collaborative—an invitation to researchers, developers, and public institutions to help prototype the future of Geo‑AI interaction. How can we make geodata more “intelligible” to agents ? Which tools, standards, or abstractions should we build together? This session is a first step toward that shared exploration.