Amazon Unveils AI Agents: ‘Kiro’ Codes Independently for Days

Amazon Web Services (AWS) has recently introduced three innovative AI agents, known as “frontier agents,” designed to enhance productivity in software development. Among these, the Kiro autonomous agent stands out for its ability to work independently for extended periods.

Kiro: The Autonomous Coding Agent

The Kiro autonomous agent builds on AWS’s earlier coding tool, Kiro, which was launched in July. This groundbreaking technology focuses on generating operational code rather than merely prototyping. Kiro ensures compliance with software-coding specifications through a method termed “spec-driven development.”

How Kiro Works

• Kiro learns from team interactions and existing code.
• It tackles complex tasks as directed, minimizing the need for constant supervision.
• It maintains performance over long periods, reportedly functioning for days before requiring human input.

During his keynote at the AWS re:Invent conference, AWS CEO Matt Garman emphasized Kiro’s ability to adapt to team preferences and standards over time. He explained that users could assign a complex task, like updating critical code for multiple applications, and Kiro could manage the entire process efficiently in one go.

Additional AI Agents Launched

AWS also unveiled two other agents to complement Kiro’s capabilities. These are:

• AWS Security Agent: This agent autonomously identifies and addresses security issues in code writing, providing suggested fixes as needed.
• DevOps Agent: It automatically tests new code for performance and compatibility with various software, hardware, and cloud solutions.

Industry Context

While AWS’s agents are not the first to promise lengthy operational periods, they mark a significant advancement in AI-driven software development. OpenAI’s recent model, GPT-5.1-Codex-Max, also claims capability for sustained tasks. However, challenges such as hallucinations and accuracy issues persist in the landscape of AI programming agents.

Developers often remain cautious, preferring to assign shorter tasks to these AI tools until they prove reliable. Nonetheless, AWS’s latest advancements indicate a positive trajectory towards integrating AI agents as valuable contributors in software development teams.