Introduction Amazon recently entered the difficult area of data context management by announcing a new set of products for AI agents. The main feature of this offering is AWS Context, which is a knowledge graph service that gets smarter based on how agents use it over time. AWS also made Amazon S3 Annotations generally available and provided a preview of skill assets within AWS Glue Data Catalog. This new AWS context layer directly challenges old methods that required people to manually maintain data graphs.
The context layer occupies a heavily debated area within computer design, because different companies offer many ways to manage organizational data. AWS enters this specific market with a unique design idea, suggesting the data graph should learn from agent usage without human review. Swami Sivasubramanian, vice president of Agentic AI at AWS, stated during the AWS Summit NYC keynote that agents could become smarter without needing to rebuild anything. He claimed this service automatically constructs a knowledge graph from all the company’s existing data sources, which simplifies the process for users.
This service figures out relationships across different data sets, business rules, and organizational knowledge, making this information available to agents and the organization while the system runs. AWS Context automatically maps relationships across current data, identifying existing tables, what columns mean, and how different sources connect. The system combines semantic search with reasoning at the graph level, thus inferring connections across various datasets and domain knowledge.
How AWS Context Builds Itself From Existing Data
AWS states that companies often struggle to manage data relationships in their current setups, which is the problem AWS context works to solve. The service automatically maps relationships across existing data, which includes identifying authoritative data sources and defining what the columns mean. It integrates semantic search with graph-level reasoning, giving agents access to complex business rules and knowledge at the point of operation.
Sivasubramanian noted that the knowledge graph improves itself as it learns which sources give correct results and which parts of data people use most often. Data stewards manage the graph through the AWS Management Console, where they can review inferred relationships and promote them to production. They also attach specific business definitions and usage rules while overseeing the graph’s evolution. Every query inherits the calling user’s IAM and Lake Formation permissions, which makes all agent data access auditable through controls that organizations already trust.
All metadata publishes in Apache Iceberg format to Amazon S3 Tables, allowing querying through Athena, Redshift, Spark, or any engine compatible with Iceberg. This design avoids proprietary APIs, which helps third-party catalog connections pull context from systems outside of AWS into the same graph. Agents query through agentic search APIs and MCP tools across Bedrock AgentCore, EKS, or any MCP-compatible framework, expanding the reach of the AWS context layer.
Building Context with Multiple Services
The field of data context is complicated, so AWS built a stack of several services to help companies build context across their entire data setup. Amazon S3 Annotations lets users attach rich business context directly to individual S3 objects at the storage level. AWS Glue Data Catalog skill assets attach specialized domain knowledge at the catalog level, linking usage rules and query patterns to data assets across the whole estate.
AWS Context synthesizes both S3 Annotations and Glue skill assets into the knowledge graph that agents query when the system is running. This process combines semantic search with graph-level reasoning across both structured and unstructured sources, ensuring each layer feeds the next. This layered approach helps organizations build a full understanding of their data environment, which is key for advanced AI agents. The overall goal of the AWS context layer is to make agents more powerful by giving them deep organizational knowledge.
The Competitive Data Context Space
AWS enters this highly competitive market where several other tech leaders already offer context solutions. Snowflake announced its context approach earlier this month using its Horizon Context and Cortex Sense services, offering a different path to data understanding. Microsoft provides context through its Fabric IQ platform, which gives a semantic structure for business data, claiming to structure knowledge for AI. Redis also developed a context platform that optimizes data specifically for retrieval, while Pinecone has its Nexus context offering that compiles data into task-specific artifacts before agents query them.
AWS’s structural argument suggests that for companies already using S3, Glue, and Lake Formation, AWS Context extends an existing identity model without needing to move data. The company claims this offers zero-integration friction, which goes beyond just consolidating costs for the user. Holger Mueller, a VP and Principal analyst at Constellation Research, told VentureBeat that context makes agents more powerful because every platform needs this capability.
Mueller cautioned that the general concern for all context offerings will involve performance, especially when handling transactional data, which remains a point of limited certainty. While AWS claims automatic learning, users should assess how quickly the graph adapts to real-world data changes and usage patterns. The presence of the AWS context layer provides organizations with a powerful, unified way to manage complex data interactions for their AI agents. For related coverage, see AI coverage.
