Build a semantic layer for agentic AI on AWS with AtScale and Amazon Bedrock AgentCore

This article demonstrates building a semantic layer for agentic AI using AtScale and Amazon Bedrock AgentCore to enable governed, business-meaningful queries over Amazon Redshift.

• AtScale models dimensional data (dimensions, hierarchies, measures) in Semantic Modeling Language (SML) YAML versioned in Git
• Strands agents query the semantic layer through AtScale's MCP server, with AgentCore Gateway managing identity propagation and credentials
• AtScale's autonomous aggregate management transparently routes queries to pre-built summary tables, accelerating agent-driven workloads
• Row-level security, perspectives, and RBAC enforce governed access at the semantic layer before queries reach the warehouse
• Audit records capture both LLM-generated dimensional SQL and warehouse-native pushdown SQL for complete traceability
• AtScale deploys on Amazon EKS using Helm charts and K8S Blueprints, connecting to Amazon Redshift Serverless

The architecture enables enterprises to answer cross-cutting business questions through agentic AI while maintaining security, governance, and query performance through a dimensional semantic model.