Amazon Web Services on Tuesday announced three new AI agents it calls “Frontier Agents.” These include agents designed to learn your preferred way of working and work alone for several days.
Each of these agents handles different tasks, such as automating code creation, security processes such as code reviews, and DevOps tasks such as preventing incidents when publishing new code. A preview version of the agent is currently available.
Perhaps AWS’s biggest and most interesting claim is its promise that its frontier agents, called “Kiro autonomous agents,” can operate independently for days at a time.
Kiro is a software coding agent based on AWS’ existing AI coding tool “Kiro” announced in July. While its existing tools can be used for vibecoding (actually just prototyping), it was intended to create production code, i.e. software that would be pushed live. To produce reliable code, AI must follow a company’s software coding specifications. Kiro accomplishes this through a concept called “spec-driven development.”
As coding, Kiro creates specifications by having humans direct, confirm, or modify assumptions. Kiro autonomous agents monitor how your team performs with different tools through training measures such as scanning existing code. And AWS says it can operate independently.
“Simply assign complex tasks from your backlog and we’ll figure out on our own how to complete that work,” AWS CEO Matt Garman promised while introducing the new product during his keynote at AWS re:Invent on Tuesday.
“We actually learn how our users want to work and continue to develop our understanding of the code, the product, and the standards our teams follow over time,” he said.
tech crunch event
san francisco
|
October 13-15, 2026
Amazon says Kiro maintains “persistent context across sessions.” In other words, you won’t run out of memory and forget what you’re supposed to do. So Amazon promises that they can be handed tasks and work independently for hours or days with minimal human intervention.
Garman described a task like updating a piece of critical code used in 15-bit enterprise software. Instead of assigning and validating each update, you can assign Kiro to fix all 15 with one prompt.
To complete the automation of coding tasks, cloud providers have developed AWS Security Agent. AWS Security Agent is an agent that works independently as you write your code to identify security issues, test them after the fact, and provide suggested fixes. DevOps agents complete these three functions, automatically testing new code for performance issues and compatibility with other software, hardware, or cloud configurations.
To be sure, Amazon agents are not the first to claim long tenures. For example, OpenAI announced last month that its agent coding model, GPT‑5.1-Codex-Max, is designed for long-running runs of up to 24 hours.
It’s also not entirely clear whether the biggest barrier to agent deployment is the context window (i.e., the ability to run continuously without outages). LLM still suffers from hallucinations and accuracy issues, leaving developers as “babysitters,” he said. Therefore, developers often want to assign short tasks and quickly validate them before proceeding.
Still, for agents to be like their colleagues, the context window needs to grow. Amazon’s technology is another big step in that direction.
Catch the latest announcements on everything from agent AI to cloud infrastructure to security and more from Amazon Web Services’ flagship event in Las Vegas. This video is provided in partnership with AWS.
