Amazon is previewing 3 AI agents, including “Kiro” that can code on its own for days

Amazon Web Services on Tuesday Announce Three new AI agents are called “boundary agents,” including one designed to learn the way you like to work and then work alone for several days.

Each of these agents handles different tasks such as writing code, security operations such as code reviews, and automating DevOps tasks such as It also prevents accidents When you pay the new code directly. Preview versions for agents are available now.

Perhaps AWS’s biggest and most interesting claim is its promise that a border agent called an “Autonomous Kiro Agent” can operate on its own for days at a time.

Kiro is a software coding agent based on AWS Kiro’s existing AI coding tool, which was… Announced in July. While this current tool can be used for dynamic coding (and is really just prototyping), it was intended to produce operational code, or programs that could be deployed directly. To create reliable code, AI must follow the company’s software coding specifications. Kero does this through a concept called “specification-driven development.”

As Kero encodes, it instructs, confirms, or corrects human assumptions, thus creating specifications. An autonomous Kiro agent monitors how the team works in different tools by scanning existing code, among other means of training. Then, AWS says, it can operate autonomously.

“You simply assign a complex task from the backlog and autonomously figure out how to get that work done,” AWS CEO Matt Jarman promised when introducing the new product during his keynote at AWS re:Invent on Tuesday.

“It’s actually learning how you like to work, and it continues to deepen its understanding of your code, your products, and the standards your team follows over time,” he said.

Amazon says Kiro maintains “continuous context across sessions.” In other words, he doesn’t run out of memory and forget what he was supposed to do. Thus, tasks can be handed off and work alone for hours or days, Amazon promises, with minimal human intervention.

Jarman described a task like updating a piece of critical code used by 15 bits of the company’s software. Instead of setting and checking each update, Kiro can be set to fix all 15 updates in one prompt.

To complement the automation of coding tasks, the cloud provider developed the AWS Security Agent, an agent that works autonomously to identify security issues as code is written, tests them after the fact, and then offers suggested fixes. A DevOps Agent completes the trio, automatically testing new code for performance issues, or compatibility with other software, hardware, or cloud settings.

Amazon agents certainly aren’t the first to demand long business windows. For example OpenAI he said last month GPT‑5.1-Codex-Max, its proxy encoding model, is also designed to run for long periods of up to 24 hours.

It’s also not entirely clear that the biggest hurdle to agent adoption is context windowing (aka the ability to run continuously without stopping). LLM owners still face hallucinations and accuracy issues that developers turn to They say “nannies.” So developers often want to assign short tasks and check them quickly before moving forward.

However, before agents can become like co-workers, context windows must grow larger. Amazon’s technology is another big step in this direction.