How to build governed agentic analytics on Amazon Redshift with the dbt Semantic Layer

Every data team has seen it: an agent confidently reports a metric, having joined the wrong tables and invented a definition no one in the business would recognize. The problem isn't the model. It's that the agent had no governed context to work from.

This session wires the dbt MCP server and the Amazon Redshift MCP Server into one architecture. Metric definitions live in the dbt Semantic Layer. MetricFlow compiles them into governed SQL. The Redshift MCP Server executes it read-only, discovering the cluster for you — so the agent never authors a query, it resolves one.

We'll demo this live: one business question in, one answer out, with a citation that traces back to the dbt model that produced it and the tests that passed on it.

Built for teams running dbt on the dbt platform against Redshift. You'll leave with three things: the reference architecture, working code you can deploy on AWS, and a clear read on which parts of your stack have to be governed before an agent touches them.