AWS Unveils Frontier AI Agents for Enterprises, Can Operate for Days Without Intervention

Amazon Web Services (AWS) unveiled frontier agents, a new class of autonomous and scalable artificial intelligence (AI) agents for enterprises. The company said that it wanted to develop AI agents that do not require constant monitoring and those which can only complete one task at a given time. There are three specialised agents in this class, which were first announced at AWS re:Invent 2025, and are aimed at software development teams. These include a coding assistant dubbed Kiro, a security analysis tool, and an operations monitoring agent.

AWS Unveils New Class of Frontier AI Agents

In a newsroom post, the cloud services provider announced its first three frontier agents, highlighting their key characteristics. These AI agents can operate autonomously and work towards a goal instead of a single task. The company claims that these are also scalable, meaning they can perform multiple tasks simultaneously and organise complex tasks across multiple agents. AWS claims these agents can also operate independently.

Coming to the three agents, first is Kiro. It is described as a virtual developer, which can manage coding tasks, navigate repositories, and maintain context over long workflows. AWS says Kiro can handle complex multi-step work, like large code refactors or cross-repository updates, without needing repeated instructions.

Second is AWS Security Agent. It is a tool that can scrutinise code for vulnerabilities, perform penetration testing simulations, and help build applications with security baked in from the start. It is said to help integrate security review into the development lifecycle, rather than as a separate post-development step.

The third agent is AWS DevOps Agent. It is an operations-oriented agent that monitors deployments and system health, helps detect and respond to incidents, and assists with ongoing reliability and performance improvements. It can analyse logs, coordinate diagnostics and recommend mitigation steps for issues.

Under the hood, these agents combine generative AI models, memory architectures and automation tooling, effectively becoming an extension of a developer or DevOps team. For instance, Kiro can clone repositories, analyse dependencies, plan tasks, submit pull requests and even run automated test suites.

While the initial launch focuses on software development workflows, AWS has hinted that these frontier agents could eventually be extended to other domains, such as data pipelines, customer support automation, infrastructure management, depending on user needs.