Quick Facts

  • AWS announced AWS Context at the AWS Summit New York City, a new service that builds a self-learning knowledge graph from existing enterprise data for AI agents.
  • The graph improves automatically as agents query it, sharing learned relationships across an organization without human re-curation.
  • AWS also announced general availability of Amazon S3 Annotations and a preview of skill assets in AWS Glue Data Catalog.

Amazon Web Services entered the enterprise context layer market with a new product designed to fix a core problem with AI agents: they often fail because they lack the organizational knowledge to reason reliably over enterprise data.

At the AWS Summit in New York City, AWS unveiled AWS Context, a service that automatically maps relationships across existing data into a knowledge graph and makes that knowledge available to AI agents at runtime. The service is listed as coming soon.

How the Graph Works

AWS Context infers what tables exist, what columns mean, how data sources relate, and which sources are authoritative. It combines semantic search with graph-level reasoning, then surfaces that knowledge through agentic search APIs and MCP tools at runtime.

The core architectural bet is that the graph should learn from agent usage rather than manual curation. As agents query the graph, AWS Context observes which sources produce correct results, which join paths get used, and which rules get applied. When one agent resolves a schema ambiguity, other agents inherit that knowledge automatically.

Swami Sivasubramanian, Vice President of Agentic AI at AWS, described the service during his Summit keynote. “Your agents now get smarter without you having to rebuild anything from scratch,” he said. “The knowledge graph improves itself over time as it learns which sources produce correct results and which parts get used.”

Governance and Access Controls

Every query through AWS Context is identity-aware. Each call inherits the calling user’s IAM and Lake Formation permissions, so an agent can only traverse relationships its identity is authorized to access. Every interaction is auditable, giving security and compliance teams a clear record of what an agent accessed and under what authority.

Data stewards manage the graph through a console, reviewing inferred relationships, promoting them to production, and attaching business definitions and usage rules. There is no infrastructure to provision and no retrieval pipeline to build.

AWS Context publishes metadata from structured and unstructured sources in Apache Iceberg format to Amazon S3 Tables, making it queryable through Amazon Athena, Amazon Redshift, Apache Spark, or any Iceberg-compatible engine. Agents built on Amazon Bedrock AgentCore, Amazon EKS, or MCP-compatible frameworks can all query it.

Built on Proven Infrastructure

The service extends the knowledge graph technology already powering Amazon QuickSight, where hundreds of thousands of users interact daily with a production graph that processes millions of requests per day. AWS Context scales that personal knowledge graph into an organizational one, shared across agents and applications.

AWS Glue Data Catalog, Amazon SageMaker Unified Studio, and AWS Lake Formation all integrate with the graph. AWS is also building connectors for third-party catalogs, so context from outside AWS can be pulled into the same graph.

A Crowded Market

AWS is not alone in this space. In 2026, Google Cloud, Microsoft, Databricks, Snowflake, MongoDB, Tableau, Redis, and Pinecone have all released tools aimed at connecting agents with enterprise context. Snowflake has its Horizon Context and Cortex Sense services. Microsoft is building context through its Fabric IQ platform. Pinecone’s Nexus offering compiles enterprise data into task-specific artifacts before agents ever query them.

AWS’s pitch to existing customers is straightforward. For enterprises already running S3, Glue, and Lake Formation, AWS Context extends an existing identity model with no data movement required.

Jake Dolezal, lead data engineer at McKnight Consulting Group, called it a meaningful shift. “It’s a meaningful shift from each team building its own retrieval-augmented generation pipeline to one governed context layer the whole organization draws from,” he said.

Read more: AWS enters the context layer race with a graph that learns from agents, not manual curation

ai agentsamazonawsdatamachine learning

This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.