LLM agents drive maps by screenshotting and clicking which is costly and blind to canvas state. The open GeoUI Intent Protocol lets agents drive geospatial viewers through typed intent ~15× cheaper per task, more accurate, and portable across different maps.
