At the AWS Summit in New York City, AWS announced a suite of new capabilities designed to provide context intelligence for data and AI agents at scale. The centerpiece of this launch is AWS Context, a service that automatically maps existing enterprise data into a knowledge graph to assist AI agents in accessing governed business rules and domain knowledge at runtime. By evolving personal knowledge graphs into a shared organizational layer, the service allows agents to draw upon cross-system relationships and curated context, with integrations for AWS Glue Data Catalog, Amazon SageMaker Unified Studio, and AWS Lake Formation.

AWS Context is designed to learn from agent behavior, ranking data sources and join paths based on usage patterns to improve performance automatically. To ensure portability, the service publishes key metadata in the Apache Iceberg format to Amazon S3, enabling customers to use any Iceberg-compatible engine for analysis or auditing. Furthermore, every query is identity-aware, inheriting the calling user’s existing IAM and Lake Formation permissions to ensure that AI agent interactions remain auditable and compliant with enterprise security standards.

Additionally, AWS announced a preview of business context and semantic search for the AWS Glue Data Catalog, which allows users to enrich data assets with glossary terms and business definitions. Alongside this, the company introduced skill assets—a new type of reference that points agents to documentation like runbooks or domain-specific process guides. This feature aims to ground agent reasoning in trusted instructions, which can be accessed via agentic search APIs or frameworks compatible with the Model Context Protocol (MCP).

Finally, AWS announced the general availability of Amazon S3 annotations, which allow developers to attach up to 1 GB of rich, mutable business context directly to S3 objects. These annotations are stored in managed Iceberg tables, making them queryable through services such as Amazon Athena or Amazon Redshift. Because annotations are bound to the objects themselves, they remain persistent during copy or replication operations, removing the need for customers to build and synchronize separate metadata databases for their AI agents.

At the AWS Summit in New York City, AWS announced a suite of new capabilities designed to provide context intelligence for data and AI agents at scale. The centerpiece of this launch is AWS Context, a service that automatically maps existing enterprise data into a knowledge graph to assist AI agents in accessing governed business rules and domain knowledge at runtime. By evolving personal knowledge graphs into a shared organizational layer, the service allows agents to draw upon cross-system relationships and curated context, with integrations for AWS Glue Data Catalog, Amazon SageMaker Unified Studio, and AWS Lake Formation.

AWS Context is designed to learn from agent behavior, ranking data sources and join paths based on usage patterns to improve performance automatically. To ensure portability, the service publishes key metadata in the Apache Iceberg format to Amazon S3, enabling customers to use any Iceberg-compatible engine for analysis or auditing. Furthermore, every query is identity-aware, inheriting the calling user’s existing IAM and Lake Formation permissions to ensure that AI agent interactions remain auditable and compliant with enterprise security standards.

Additionally, AWS announced a preview of business context and semantic search for the AWS Glue Data Catalog, which allows users to enrich data assets with glossary terms and business definitions. Alongside this, the company introduced skill assets—a new type of reference that points agents to documentation like runbooks or domain-specific process guides. This feature aims to ground agent reasoning in trusted instructions, which can be accessed via agentic search APIs or frameworks compatible with the Model Context Protocol (MCP).

Finally, AWS announced the general availability of Amazon S3 annotations, which allow developers to attach up to 1 GB of rich, mutable business context directly to S3 objects. These annotations are stored in managed Iceberg tables, making them queryable through services such as Amazon Athena or Amazon Redshift. Because annotations are bound to the objects themselves, they remain persistent during copy or replication operations, removing the need for customers to build and synchronize separate metadata databases for their AI agents.