ktx-ai-data-context-layer

To build a semantic layer for AI agents, you need to ingest metadata from databases, dbt, Looker, Metabase, and wikis into a centralized context layer. This provides a unified source of truth for schemas and business rules that agents can query via MCP tools.

Providing dbt and Looker context to an MCP-driven agent requires ingesting their metadata to map joins and capture usage rules. This Skill assembles that metadata into a semantic layer, exposing semantic entities directly to agents for accurate data querying.

You set up a project structure for metadata ingestion by initializing the context layer with a ktx.yaml configuration file. This creates dedicated semantic-layer and wiki directories to organize metadata for MCP-driven agent inquiries.

Yes, you can use this semantic layer with Metabase and dbt simultaneously. It ingests metadata from both platforms alongside databases and wikis, mapping joins and capturing usage to expose a unified semantic context for your AI agents.

AI agents lack context about business rules because they do not have a single source of truth for schemas and joins. Ingesting metadata from wikis and BI tools into a centralized semantic layer resolves this by providing accurate, unified context for agent inquiries.