Amazon unveils three AI assistants, featuring 'Kiro' that is capable of independently writing code for several days

AWS Unveils Advanced AI Agents for Automation

On Tuesday, Amazon Web Services introduced three innovative AI agents, referred to as “frontier agents.” Among them is a tool designed to adapt to your workflow and operate autonomously for extended periods.

These new agents are tailored for distinct roles, including code generation, security oversight such as code reviews, and automating DevOps responsibilities like incident prevention during code deployment. Early access to these agents is now available for users to try.

Kiro Autonomous Agent: Long-Term Independent Coding

The most notable announcement centers on the “Kiro autonomous agent,” which AWS claims can function independently for several days at a time.

Kiro builds upon AWS’s previously launched AI coding assistant, first revealed in July. While the original tool was aimed at rapid prototyping and generating production-ready code, the new Kiro agent is engineered to adhere to organizational coding standards through a method called “spec-driven development.”

As Kiro writes code, it interacts with human team members to clarify, validate, or adjust its assumptions, thereby forming detailed specifications. The autonomous agent learns by observing team workflows and analyzing existing codebases, allowing it to eventually operate without ongoing supervision.

“You can assign a complex task from your backlog, and it will independently determine how to accomplish it,” AWS CEO Matt Garman explained during his keynote at AWS re:Invent.

Garman further noted, “Kiro adapts to your preferred working style, continuously deepening its understanding of your codebase, products, and team standards.”

According to Amazon, Kiro retains “persistent context across sessions,” meaning it remembers its objectives and can tackle assignments for hours or even days with little human input.

For example, if a critical code update is needed across 15 different software components, Kiro can be tasked to handle all updates in a single instruction, eliminating the need for manual oversight of each change.

Expanding Automation: Security and DevOps Agents

To further streamline software development, AWS also introduced the Security Agent, which autonomously detects vulnerabilities as code is written, conducts post-development testing, and suggests remediation steps. The DevOps Agent complements this by automatically testing new code for performance, compatibility, and integration with various environments.

Industry Context and Ongoing Challenges

While Amazon’s agents promise extended autonomous operation, they are not the first to do so. For example, OpenAI recently announced that its GPT‑5.1-Codex-Max model is also designed for long-duration tasks, capable of running up to 24 hours.

However, the main barrier to widespread adoption of autonomous agents may not be their ability to work continuously. Developers still face challenges with large language models, such as hallucinations and accuracy issues, which often require frequent human oversight. As a result, many developers prefer to assign smaller, easily verifiable tasks.

Nonetheless, expanding the context window is essential for agents to function more like true collaborators. Amazon’s latest advancements mark a significant move toward that goal.