Amazon previews 3 AI agents, including ‘Kiro’ that can code on its own for days

In a groundbreaking announcement, Amazon Web Services (AWS) introduced three innovative AI agents, collectively known as 'Frontier agents.' Among them is ‘Kiro,’ a remarkable autonomous coding agent designed to learn from user preferences and operate independently for extended periods. These AI agents are engineered to tackle a range of tasks, including writing code, conducting security reviews, and automating DevOps processes to prevent incidents during code deployment. The preview versions of these agents are now accessible to users. Kiro, the standout of the trio, is built on AWS's existing AI coding tool, also named Kiro, which was previously revealed in July. While the earlier version was primarily aimed at prototyping, Kiro is now equipped to generate operational code that can be deployed live. This is achieved through a method called 'spec-driven development,' where Kiro collaborates with human developers to create and refine coding specifications. As Kiro learns from the team’s workflows and analyzes existing code, it is said to gain the ability to function autonomously. AWS CEO Matt Garman highlighted this capability during his keynote at AWS re:Invent, stating, "You simply assign a complex task from the backlog and it independently figures out how to get that work done." Over time, Kiro deepens its understanding of the project’s coding standards and the team’s preferences. One of the notable features of Kiro is its ability to maintain persistent context across sessions, allowing it to remember tasks without losing track of its objectives. This means it can take on assignments and work independently for hours or even days with minimal human oversight. For example, Garman illustrated how Kiro could be tasked with updating critical code across multiple corporate software applications, streamlining the process by handling all updates with a single prompt. In addition to Kiro, AWS also rolled out the AWS Security Agent, which autonomously identifies security vulnerabilities in code and suggests fixes, and the DevOps Agent, which tests new code for performance and compatibility issues. While Amazon’s AI agents are not the first to promise long operational capabilities, as OpenAI recently announced similar features for its GPT-5.1-Codex-Max model, the challenges of context awareness and accuracy remain. Developers often prefer to assign shorter tasks to mitigate the risk of AI errors, as many still face issues with hallucinations and inconsistencies in their outputs. Nonetheless, Amazon's advancements signify a substantial step forward in the evolution of AI technology, moving closer to the vision of AI functioning as indispensable team members in coding environments.