AWS launches Context service to give AI agents intelligent data awareness at scale

AWS Context is a new service launching soon at the AWS Summit New York City. It automatically builds a knowledge graph that maps relationships across your existing data stores—including data lakes, warehouses, lakehouses, databases, and streams. AI agents can then query this graph through agentic search APIs and MCP tools, retrieving governed data relationships, business rules, and domain knowledge at runtime. Data stewards manage the graph via an intuitive console, reviewing inferred relationships and attaching business definitions. The technology extends the same knowledge graph used by Amazon Quick, which handles millions of daily requests. AWS Context transforms that personal graph into an organizational shared context layer, giving Quick’s agents immediate access to enterprise-wide relationships and curated rules. Integration with AWS Glue Data Catalog, SageMaker Unified Studio, and Lake Formation allows teams to govern with business rules and permissions, and new context can be added via AI assistance or manual curation.

The service continuously learns from agent usage: it observes which sources produce correct results, which join paths agents rely on, and which curated rules are applied, then shares these insights across the organization. All key metadata is published to Amazon S3 in Apache Iceberg format, making it queryable via Athena, Redshift, Spark, or any Iceberg-compatible engine. This ensures the context is portable and not locked into AWS. AWS Context is also designed to connect to third-party catalogs, so datasets from outside AWS can be brought into the same graph. Critically, every query is identity-aware, inheriting the calling user’s IAM and Lake Formation permissions. This means an agent can only see and traverse relationships its identity is authorized to access, providing full governance and auditability from day one.