AtScale semantic layer bridges agents and BI on Amazon Bedrock AgentCore

This is the third entry in a series that has kept the same Customer 360 request while swapping the semantic layer underneath it — first Stardog's knowledge graph, then Denodo's data virtualization, now AtScale's dimensional model. The shift is architectural: Stardog and Denodo federate at query time across engines, whereas AtScale plans a dimensional query as native warehouse SQL and pushes it to one engine per query. In this walkthrough, all analytical C360 data lives in Amazon Redshift Serverless, with Amazon Aurora PostgreSQL kept as the operational system of record and a replication path keeping the analytical copy current.

The trade-off is explicit in the post. Consolidating into one engine rules out the live cross-engine federation the earlier two layers perform, but it lets AtScale's planner and autonomous aggregate management work against a single-engine execution model. That matters for agent-driven workloads, which fire more queries than a human analyst would; matching queries get routed to pre-built summary tables in the warehouse rather than rescanning the fact table.

Governance is framed as identical across consumers: one row-security object, one perspective, and one set of role bindings apply whether the caller is the agent over MCP, Excel over XMLA, or a notebook over JDBC. The series' two prior posts did not capture both the LLM-written query and the warehouse-executed query in a single record, which the post presents as the audit advantage here.