Context intelligence for your data and AI agents at scale

Agents are only as intelligent as the context they can reason over. Today, that context is scattered across data lakes, data warehouses, lakehouses, databases, and streams, and in institutional knowledge that has never been written down. You want to trust the decisions made by your AI agents, but that can’t happen until agents have context. Imagine what becomes possible when we give agents a safe way to access the context they need to deliver trusted decisions.

This is why at the AWS Summit New York City, we’re announcing a series of innovations that deliver context intelligence for your data and AI agents at scale. 

AWS Context (Coming soon)

In today’s keynote, we introduced AWS Context, a new service that automatically maps the relationships across your existing data into a knowledge graph and provides agentic search so AI agents in the organization can access governed data relationships, business rules, and domain knowledge at runtime. Data stewards and curators manage the graph through an intuitive console experience, reviewing inferred relationships, promoting them to production, and attaching domain-specific knowledge like business definitions and usage rules. 

AWS Context extends the same knowledge graph technology that powers Amazon Quick, where hundreds of thousands of users interact daily with a production knowledge graph that catalogs datasets, dashboards, and metadata, learning from usage patterns to make every interaction smarter. That graph already processes millions of requests per day. With AWS Context, we are extending what was a personal knowledge graph into an organizational one, a shared, governed context layer that agents and applications in your organization can draw from. Existing Amazon Quick users benefit immediately. When AWS Context is enabled, Quick’s agents gain access to the broader enterprise knowledge graph, including cross-system relationships, business rules, and curated context that go beyond what any single user’s personal graph can provide. AWS Glue Data Catalog, Amazon SageMaker Unified Studio, and AWS Lake Formation integrate with the knowledge graph, so teams can govern it with business rules and permissions and add new context automatically with AI assistance or explicitly through manual curation. 

Key elements of the context layer are published to Amazon S3 in the Apache Iceberg format, so that customers are free to use the Iceberg-compliant tools of their choice to consume metadata and build against AWS Context based on open standards. There is no infrastructure to provision or retrieval pipeline to build, and customers can begin to gather and curate context for their agents with just a few clicks in the AWS Management Console. 

Let’s take a closer look at the capabilities behind it. 

Context that learns from how your agents work

AWS Context gets smarter the more your agents use it. As agents query the graph, it observes which sources produce correct results, which join paths agents rely on, and which curated rules get applied. It ranks sources by actual usage and shares what it learns across your organization, so when one agent discovers a correct join path or resolves a schema ambiguity, other agents pick it up, without requiring a human re-curate the graph. 

Open and portable by design

AWS Context publishes all key metadata from structured and unstructured sources into Apache Iceberg format in Amazon S3 Tables, so you can query your context with Amazon Athena, Amazon Redshift, Apache Spark, or any Iceberg-compatible engine, and build downstream systems on it, audit it, or migrate it. 

AWS Context is also designed to connect to third-party catalogs, so you can bring context from systems beyond AWS into the same graph. Agents query it through agentic search APIs and MCP tools, whether they’re built on Amazon Bedrock AgentCore, deployed on Amazon EKS, or running on MCP-compatible frameworks. Your context stays queryable, portable via Apache Iceberg format, and fully yours. 

Identity-aware and governed by default

Any agent you put into production raises a governance question: what data can it reach, and can you show exactly what it accessed and under whose authority? AWS Context answers both by making every query identity-aware. Each call is designed to inherit the calling user’s IAM and Lake Formation permissions, so an agent can only see and traverse the relationships its identity is authorized to access. Because access runs through identity, every interaction is auditable. Your security and compliance teams can verify what an agent accessed and under what authority, using the same controls you already rely on. 

AWS Glue Data Catalog Business Context and Semantic Search (preview)

Today, we also announced the preview of business context and semantic search for AWS Glue Data Catalog, providing context and tools that makes it easier for humans and AI agents to discover and understand data. Customers can now enrich their Glue tables, views, and columns, including those backed by S3 Tables, with business descriptions, glossary terms, custom metadata, and associate them with skill assets that provide additional data context stored outside the catalog. With business context indexed alongside technical metadata in Glue Data Catalog, customers can use the new Glue Search API to more quickly find data by business meaning and AI agents can ground their reasoning in trusted definitions rather than inferring context. 

We are also excited to offer a preview of skill assets in Glue Data Catalog. Now, data producers can create skill assets, a new asset type that references URIs to files (such as AI skills, guide markdown files, and team runbooks) hosted in any location including S3, git repositories, and wikis. Associating skill assets to data assets gives agents additional context and instructions they can retrieve progressively for working with specific data without re-teaching it to every agent one prompt at a time. For example, the skill URI locations can point to your team’s repositories with domain-specific documentation or processes that include data usage details such as the grain and scope, common query patterns and best practices, and usage rules (when to use the data, what are join keys and required filters).  

Skill assets make it easier for AI agents to find the right data to use in a data estate but that is only half the problem. An agent also has to know how to use it: the filters to apply before aggregating data, the join paths to follow, the caveats that aren’t visible in the technical schema. Today, the AWS Agent Toolkit contains default skills to help AI agents work with Glue Data Catalog as well as other capabilities like Amazon Athena and S3 Tables. Many enterprises have their own skills that their data teams have developed. To get started, developers can connect any MCP-compatible agents using the remote, fully managed AWS MCP to access AWS service skills or by installing the aws-data-analytics plugin for Claude Code, Cursor, and Amazon Kiro, to ask an agent to find data, conduct analysis, or build applications on top of that data using AWS or other custom skills. Agents built with AgentCore harness can access all AWS skills in the AWS Agent Toolkit with one line of code. This enables your agents to take advantage of the AWS service expertise and best practices quickly.  

Amazon S3 Annotations (generally available)

To make it easier for customers to add their own custom context to their data lake, we announced the general availability of Amazon S3 annotations, a new way to attach rich, queryable business context directly to your S3 objects and store that context in an S3 Iceberg table. Customers have long described their objects in S3 with object tags and user-defined metadata, and those remain the right tools for operational tasks like access control and for small pieces of information set at upload. But as customers build agents on top of their data, they want to attach far more metadata. They want to create and evolve rich context that an agent can read and act on, at scale. S3 annotations provide that capability in an open data format. Each object stored in S3 can have up to 1 GB of context. Annotations are mutable, so you can evolve context as your data changes. S3 annotations live with the S3 object in S3 storage. That means that S3 annotations move with its associated S3 object through copy and replication operations, and they’re removed when the object is deleted. With annotations, there’s no separate metadata database to build, synchronize, or keep from drifting out of date. 

Annotations become queryable through S3 Metadata. When you enable annotation tables on a bucket, every annotation flows automatically into a fully managed Iceberg table. You can query across all your objects with Amazon Athena, Amazon Redshift or any Iceberg-compatible engine, and agents can discover annotations in natural language through the S3 Tables MCP server. 

With Amazon S3 annotations, you attach rich business context directly to S3 objects and query it at scale, so agents find what they need without building separate metadata systems. 

Context is the data lake for AI agents, and with these innovations, we are building the foundation of knowledge and intelligence for AI agents interacting with data across organizations and enterprises of any scale.

About the author

Mai-Lan Tomsen Bukovec

Mai-Lan Tomsen Bukovec, Technology Vice President at AWS, leads the Amazon cloud data services that millions of AWS customers rely on for digital transformations, business analytics, machine learning, generative AI, and next generation customer experiences. With over 25 years of experience in the technology industry, Mai-Lan is a pioneer in helping customers take advantage of cloud-based technologies to transform their businesses.