Agents are as intelligent as the context they can consult with each other. Today, that context is spread across data lakes, data warehouses, lake housesdatabase, streams, and in Institutional information that has never been written down. You want to trust the decisions made by your AI agents, but that I won't occurs until agents have context. Imagine what could happen if we gave agents a secure way to access the context they need to deliver trusted decisions.
That's why at the AWS Conference in New York City, we are here announceI am a series of new that deliver intelligence to your data and AI agents at scale.
AWS Context (Coming Soon)
In today's keynote, we introduced AWS Context, a new service that automatically creates relationships across all your existing data in a knowledge graph and provides agency search so AIs in the organization can access governing data relationships, business rules, and domain information at runtime. Data managers and administrators manage the graph using an intuitive console experience, review specific relationships, escalate to production, and attach domain-specific information such as business definitions and usage rules.
AWS Context extends the same information graph technology that powers Amazon Quick, where hundreds of thousands of users interact daily with a production information graph that includes datasets, dashboards, and metadata, learning from usage patterns to make every interaction smarter. That graph is already processing millions of requests per day. With AWS Context, we extend what was once a personal information graph into an organizational one, a shared, managed context layer that agents and applications in your organization can draw from. Existing Amazon Quick users benefit immediately. With AWS Context enabled, Quick agents gain access to a broader business information graph, including cross-system relationships, business rules, and selective context beyond what a single user's personal graph can provide. AWS Glue Data Catalog, Amazon SageMaker Unified Studio, and AWS Lake Formation integrate the information graph, so that teams can manage it with business rules and permissions and add new context automatically with the help of AI or clearly with manual processing.
Key features of the context layer are published on Amazon S3 in Apache Iceberg format, so customers are free to use their favorite Iceberg-compliant tools to use metadata and build against AWS Context based on open standards. There is no provisioning or retrieval pipeline infrastructure to build, and customers can start collecting and managing content for their agents with just a few clicks in the AWS Management Console.
Let's take a closer look at the capabilities behind it.
The context you learn from how your agents work
AWS context it becomes smarter if your agents use it a lot. As agents query the graph, it you notice which sources produce the relevant results, which methods agents rely on, and which selection rules are used. It measures sources with actual usage and shares what it learns across your organization, so when one agent finds the correct join method or resolves a schema ambiguity, another agent.s pick it up, withoutside he needs a of a person reorganizee graph.
It is open and portable in design
AWS Context publishes all key metadata from structured and unstructured sources to 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 subsystems on it, test 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 inquire about agent search APIs and MCP tools, whether they are built on Amazon Bedrock AgentCore, implemented on Amazon EKS, or use MCP-compatible frameworks. Your content is always on demand, portable in Apache Iceberg format, and fully yours.
It is Identity-aware and automatically managed
Any agent you put into production raises the question of governance: what data does it have access to, and can you show exactly what it has accessed and under whose authority? AWS content answers both with to do all the questions self-awareness. Each call designed to do get the user's IAM and Lake Formation permissions, so the agent can only see and terminate relationships that it owns is authorized to access. Because access is identity-based, all interactions are readable. Your security and compliance teams can verify what an agent has 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 a preview of business context and semantic search for the AWS Glue Data Catalog, providing context and tools that make it easier for humans and AI agents to discover and understand data. Customers can now enrich their Glue tables, views, and columns, including those supported by S3 Tables, with entity definitions, glossary names, custom metadata, and associate them with capability properties that provide context for additional data stored outside the catalog. With business context embedded alongside technical metadata in the Glue Data Catalog, customers can use the Glue Search API to quickly find data by business definition and AI agents can focus their thinking on trusted definitions rather than context.
We are also pleased to offer a preview of the capabilities in the Glue data catalog. Now, data producers can create capabilities assets, a new type of asset that points to URIs to files (such as AI capabilities, markdown directive files, and team runbooks) hosted in any environment including S3, git repositories, and wikis. Associating skill properties with data assets gives agents more context and instructions that they can continuously access to work with specific data without having to reteach it to every agent on the fly. For example, the ability to URI fields can point to your group's collections of domain-specific documents or processes that include data usage information such as this. such as grain and scope, common query patterns and best practices, and usage rules (when to use the data, what are the join keys and filters needed).
Capability assets make it easier for AI agents to find the right data to use in a data environment but that's only part of the problem. The agent must also know how to use it: the filters to be applied before combining the data, the integration methods to be followed, the hidden caveats in the technical schema. Today, the AWS Agent Toolkit contains automated capabilities to help AI agents work with Glue Data Catalog and other capabilities such as Amazon Athena and S3 Tables. Many businesses have their skills developed through their data teams. To get started, developers can connect anywhere MCP-enabled agents using a remote, fully managed AWS MCP to access AWS service capabilities or by installing the aws-data-analytics plugin for Claude Code, Cursor, and Amazon Kiro, to request the agent to obtain data, perform analytics, or build applications on top of that data using AWS or other custom capabilities. Agents built with the AgentCore harness can access all AWS capabilities in the AWS Agent Toolkit with a single line of code. This allows your agents to quickly adopt AWS service technologies and best practices.
Amazon S3 Annotations (often available)
To make it easier for customers to add their own custom context to their data pool, we announced the general availability of Amazon S3 Annotations, a new way to attach rich, queryable business content directly to your S3 objects and store that context in an S3 Iceberg table. Customers have long defined their objects in S3 with object tags and user-defined metadata, and those remain the ideal tools for operational tasks such as access control and small pieces of information set on upload. But as customers build agents on top of their data, they want to attach more metadata. They want to create and transform rich context that an agent can learn and act on, at scale. S3 Annotations provides that capability in an open data format. Each object stored in S3 can contain up to 1 GB of content. Annotations are scalable, so you can change context as your data changes. S3 annotations reside with the S3 object in S3 storage. That means that S3 annotations move with their associated S3 object through copy and duplicate operations, and are deleted when the object is deleted. With annotations, no metadata database can be created, synchronized, or avoided from becoming obsolete.
Annotations can be queried via S3 Metadata. When you enable annotation tables in a bucket, all annotations automatically flow into a fully managed Iceberg table. You can query all your objects through Amazon Athena, Amazon Redshift or any Iceberg-compatible engine, and agents can get annotations in natural language through the S3 Tables MCP server.
With Amazon S3 annotationsyou attach rich business content directly to S3 objects and query at scale, so agents can find what they need without building separate metadata systems.
The core is the data pool for AI agents, and with these new processes, we are building the knowledge base and intelligence of AI agents that interact with data across organizations and businesses of any scale.
About the author
Mai-Lan Tomsen Bukovec
Mai-Lan Tomsen Bukovec, Vice President of Technology at AWS, leads the Amazon cloud data services that millions of AWS customers rely on for digital transformation, business analytics, machine learning, productive AI, and the next generation customer experience. With over 25 years of experience in the technology industry, Mai-Lan is a pioneer in helping clients take advantage of cloud-based technology to transform their businesses.
