ktx-data-agent-context-layer

To stop AI agents from generating incorrect SQL, build a semantic context layer that grounds agent queries in approved metric definitions and business logic derived from dbt artifacts and warehouse metadata.

A semantic layer for AI agents provides canonical metrics, dimensions, and relationships built by introspecting your data warehouse and ingesting dbt models, exposing them via read-only MCP tools for accurate SQL generation.

You can connect wiki documentation to data warehouse metrics for AI agents by configuring context ingestion from wiki sources, dbt docs, and warehouse metadata to resolve terminology and business definitions into a unified semantic layer.

Yes, you can use MCP tools to search for metric definitions before generating SQL, as this Skill exposes read-only MCP tools for semantic search and wiki retrieval to help agents select canonical logic.

No, the semantic layer does not write back to the data warehouse during agent queries; it enforces a read-only guardrail to reduce risk and prevent modifications to your underlying data infrastructure.

To set up a dbt semantic layer for analytics agents, run the project setup command in your analytics directory, configure warehouse and context sources, ingest metadata to build the layer, then start the MCP server.