AWS used its Summit in New York to unveil two services aimed at making AI agents more useful in business settings. The company introduced AWS Continuum, a security-focused service for code vulnerabilities, and AWS Context, a knowledge graph designed to give agents better access to company data and rules.
The announcements reflect AWS's view that the next phase of enterprise AI is less about experimentation and more about putting agents to work across teams, systems and workflows. In a blog post tied to the event, Swami Sivasubramanian, AWS vice president of agentic AI, said companies are moving from talking about agents to using them to build apps, secure systems, answer customer questions and make decisions autonomously.
AWS described Continuum as an AI-native security service that continuously finds vulnerabilities, checks whether they are actually exploitable, ranks them using business context and helps fix them across the stack. The service is built to work with different models and is intended to provide visibility into why it recommends a particular action. AWS said every step is explainable and auditable, and that the system feeds results back into future cycles. The company also said Continuum includes a threat modeling feature that can generate threat models from design documents or source code.
Context is meant to solve a different problem. AWS said the service automatically creates a knowledge graph from a customer's existing data and infers how assets, business rules and domain knowledge relate to one another. The goal is to help agents find the right information faster and produce better answers. AWS said Context can connect to data across databases, documents, email, Slack and other sources, while enforcing governance so agents only see information they are allowed to access.
The company positioned Context as an agentic search layer that can be used across an organization. AWS said it is built on the same knowledge graph technology used by Amazon Quick, and that metadata can be stored in Iceberg format in S3 Tables. AWS also said the service does not require customers to provision infrastructure or build their own retrieval pipeline.
AWS framed both launches as part of a wider set of agent-related updates across its product line. Amazon Quick now has autonomous agents that can work in the background, while Kiro, AWS's software development agent, is now available on iOS so users can manage projects from a phone. AWS DevOps Agent is also gaining release management features that can review code changes, generate test plans and help catch problems before production.
Together, the announcements show AWS's effort to build more of the infrastructure required for enterprise agents to operate at scale. That includes not only coding tools, but also systems for security, data access, release management and workflow orchestration.
AWS said the new services are designed to help organizations build trust in agents by giving them the context and safeguards needed to take on more responsibility. The company also highlighted the growing importance of speed, arguing that companies using agents more effectively will be able to compound their advantage as those systems learn from repeated interactions and improve over time.