AWS Launches Context: Self-Learning Knowledge Graph for AI Agents That Gets Smarter Through Use, No Manual Re-Curation Required

Amazon unveiled AWS Context at AWS Summit New York (June 17), a knowledge graph service that automatically maps relationships across existing enterprise data—tables, columns, business rules, and domain knowledge—and exposes them to AI agents at runtime without requiring data migration. Its key architectural differentiator: the graph learns from how agents actually use it, observing which sources produce correct answers and re-ranking them so that every agent in the organization benefits from what other agents discover, eliminating the manual re-curation loop that hampers rival graph products. AWS Context launches alongside Amazon S3 Annotations (now GA) and AWS Glue Data Catalog skill assets (preview), forming what AWS VP Swami Sivasubramanian described as an integrated context intelligence stack for agentic workloads built on open Apache Iceberg.