The AWS Machine Learning Blog describes how to construct a governed semantic layer for agentic AI within EDB Postgres AI and access it from a Strands Agents agent running on Amazon Bedrock AgentCore. In this approach, business meaning, join paths, and access control are stored and managed in Postgres itself. Permissions are enforced using Postgres mechanisms such as GRANT statements and row-level security, which determine what data each agent is allowed to read.
Why it matters
Placing the semantic layer and access controls directly in the database keeps governance close to the data, allowing agents to query business information while respecting defined read permissions.
Who should care
Teams working with agentic AI on AWS, as well as developers and data practitioners using EDB Postgres AI who need governed, access-controlled data access for agents.