Amazon Web Services (AWS) announced the introduction of three new artificial intelligence (AI) agents, termed "frontier agents," during its re:Invent conference on Tuesday. These agents are designed to automate various industrial and technical processes, with a particular focus on software development, security, and operations.
Among the new offerings is the Kiro autonomous agent, which AWS claims can operate independently for several days by learning a team's specific coding practices and standards. This agent builds upon AWS's existing Kiro AI coding tool, launched in July, which focuses on "spec-driven development" to produce operational code tailored to company specifications.
AWS CEO Matt Garman, speaking at the re:Invent keynote, stated that users can assign complex tasks to the Kiro autonomous agent from a backlog, and the agent will independently determine how to complete the work. Garman added that the agent "actually learns how you like to work, and it continues to deepen its understanding of your code and your products and the standards that your team follows over time." The Kiro autonomous agent is designed to maintain "persistent context across sessions," enabling it to handle tasks requiring continuous operation over extended periods with minimal human intervention, according to AWS.
In addition to Kiro, AWS introduced two other frontier agents. The AWS Security Agent is designed to autonomously identify security vulnerabilities during code writing, conduct post-development tests, and propose corrective actions. The AWS DevOps Agent completes the trio, automating the testing of new code for performance issues and ensuring compatibility with existing software, hardware, or cloud environments.
These developments align with a broader industry trend toward AI agents with extended operational capabilities. OpenAI previously announced its GPT-5.1-Codex-Max agentic coding model, which is designed for continuous runs up to 24 hours. While the expanding "context windows"—the ability for AI models to maintain continuous understanding and memory over time—are considered a significant step forward, industry discussions continue regarding challenges such as AI hallucinations and accuracy issues that can necessitate human oversight, according to sources familiar with current developer experiences.
