AWS launched a new Context Service and significantly expanded its AgentCore platform at its annual summit in New York, marking a major step in building enterprise-grade infrastructure for agentic AI systems. The announcements aim to give AI agents governed, secure access to organizational data while providing managed tools for production deployment.
The centerpiece of the launch is AWS Context, a service coming soon that automatically constructs a knowledge graph from an organization’s existing data lakes, warehouses, databases, and streams. This unified graph allows AI agents to query relationships across enterprise information at runtime. The technology extends knowledge graph capabilities already used inside Amazon Quick, which handles millions of daily requests from hundreds of thousands of users. Every query through AWS Context inherits the calling user’s IAM and Lake Formation permissions, ensuring agents access only authorized data. The service publishes metadata in Apache Iceberg format to Amazon S3, enabling customers to query the context layer with Amazon Athena or Amazon Redshift.
Alongside Context, AWS announced that Amazon Bedrock Managed Knowledge Base is now generally available. This fully managed retrieval-augmented generation service automates the end-to-end RAG pipeline with native data connectors and smart parsing. Web Search on AgentCore also reached general availability, grounding agent responses in current web information from an index of tens of billions of documents built by Amazon. Queries are processed entirely within AWS without sending data to external search engines.
AgentCore’s optimization capabilities expanded with generally available batch evaluations, recommendations, and A/B tests. Failure, intent, and trajectory insights entered preview. On the security front, SentinelOne announced an integration delivering runtime guardrails for AI agents through the AgentCore policy engine. Rubrik disclosed a similar integration for monitoring, governance, and remediation across multi-cloud environments.
The summit also featured general availability for Amazon S3 Annotations, letting customers attach up to one gigabyte of queryable business context directly to S3 objects. A preview of business context and semantic search for AWS Glue Data Catalog was announced. These releases collectively represent AWS’s push to create what the company calls “the data lake for AI agents” a governed infrastructure layer where enterprise context is accessible to automated systems at scale.
Read Article: Actis Joins Investor Sell-Off Trend in Pine Labs with Rs 151.6 Crore Deal
