Enabling AI Agents for Semantic 3D City Models through Automated Domain Context Generation

Abstract. Semantic 3D city models offer rich urban information yet remain largely inaccessible to non-expert users due to their complex schemas, deep class hierarchies, and heterogeneous data structures. Large Language Models (LLMs) have demonstrated remarkable capabilities in natural language understanding and code generation. But applying them to domain-specific databases requires contextual knowledge that far exceeds what can be manually written and maintained. This paper presents a framework that automatically generates domain context from CityGML data stored in 3DCityDB instances. The framework decomposes domain knowledge into static components (database schema, query patterns, spatial capabilities) and dynamic components (available object classes, properties and generic attribute classifications). These components are assembled into a structured context representation served via the Model Context Protocol (MCP), forming a shared knowledge layer that most AI agents can consume. We evaluate the framework against five CityGML datasets from different countries where attribute values are given in different languages, levels of detail, and thematic modules. Results demonstrate that the system enables an LLM-based agent to correctly formulate complex SQL queries (including 3D spatial operations, multi-level feature relationships, and nested property access). Beyond querying, the generated context supports the creation of additional domain-specific agents for tasks such as semantic enrichment, data quality assessment, and urban scenario analysis, making the MCP Server a reusable domain knowledge interface for semantic 3D city models.