AWS has introduced three new AI “frontier agents” that are intended to act less like tools and more like additional members of a software team. The company says these agents can operate autonomously for long stretches, distribute tasks across multiple workflows and function without constant supervision, with the aim of reducing the bottlenecks that slow development.
AWS arrived at this approach after analysing its own internal engineering processes. It says teams improved output when agents were directed toward broad goals rather than narrowly defined tasks. It also found that productivity increased when multiple agentic tasks were run in parallel, and when agents could work for hours or days without human intervention. These insights led to a new class of AI agents that are autonomous, scalable and capable of extended independent operation.
The first of the new products is Kiro, described as a virtual developer that keeps work moving in the background. AWS says it maintains persistent context across sessions and learns from pull requests and feedback. Kiro can triage bugs, improve code coverage and handle changes that span multiple repositories. It connects to tools like Jira, GitHub and Slack so that its understanding of a team’s codebase deepens as work progresses.
The second agent, AWS Security Agent, focuses on strengthening security from early design through deployment. It reviews design documents, scans pull requests against an organisation’s security requirements and carries out on-demand penetration testing. SmugMug staff software engineer Andres Ruiz says: “AWS Security Agent helped catch a business logic bug that no existing tools would have caught, exposing information improperly.” Ruiz says the agent’s ability to contextualise information and identify unexpected API responses “represents a leap forward in automated security testing”.
The third agent, AWS DevOps Agent, is designed to respond immediately when incidents occur, isolating issues and identifying root causes. It maps application resources and correlates telemetry from observability tools with code and deployment data to reduce investigation times.
Jason Sandery, head of cloud services at Commonwealth Bank of Australia, says: “AWS DevOps Agent thinks and acts like a seasoned DevOps engineer, helping our engineers build a banking infrastructure that’s faster, more resilient, and designed to deliver better experiences for our customers.”
AWS positions the trio as a step toward more sophisticated automation across the software lifecycle. It argues that earlier generations of AI tools offered useful help on individual tasks but still required humans to reconnect context scattered across bug trackers, repositories and chat systems. The frontier agents are described as capable of taking on extended, goal-driven projects in a manner closer to how teams already work.
The launch marks an attempt to turn AI into a constant presence throughout development, security and operations, rather than a set of isolated shortcuts. AWS says the intent is to help teams spend less time on repetitive coordination and more time on the decisions that shape their products.