At re:Invent 2025, AWS unveiled a major innovation that will revolutionize the software development lifecycle – Frontier Agents. This new category of AI agents is designed to be autonomous, scalable, and work independently, enabling them to perform complex tasks for hours or even days, just like real team members, without constant human intervention.
AWS has launched three "Frontier Agents" in its first wave, targeting three core areas of need: development, security, and operations.
Kiro Autonomous Agent: Virtual Developer
Based on the previously launched AI agent serviceKiro, an integrated development environmentBased on this, Kiro is positioned as a "virtual developer" that can resolve friction caused by frequent task switching and cross-repository coordination during the development process.
• Independent work:Developers can directly assign tasks to Kiro on GitHub (such as fixing bugs or improving code coverage). Kiro will autonomously figure out how to complete the task, make changes across multiple repositories, and finally submit the results as a PR (Pull Request) for review.
• Continuous learning:Kiro maintains the context across conversations and learns from every code review and feedback. Over time, Kiro's understanding of the team's codebase and standards will deepen.
AWS Security Agent: Virtual security engineer, penetration testing completed in hours.
To address cybersecurity challenges, AWS has launched Security Agent, which not only proactively reviews design documents and code to ensure they comply with organizational security standards, but also transforms time-consuming penetration testing into on-demand capabilities.
SmugMug, a web-based photo storage database service company, shared its testing experience, pointing out that the agent can discover business logic vulnerabilities that traditional tools cannot detect, and can shorten penetration testing time from several days to several hours, thus significantly reducing costs.
AWS DevOps Agent: A virtual operations team member, enabling lightning-fast root cause analysis.
When a system failure occurs, the DevOps Agent is always ready. It integrates observation tools and code library knowledge such as CloudWatch, Datadog, and Splunk, and can automatically correlate telemetry data and deployment information to accurately locate the root cause of the problem.
The Commonwealth Bank of Australia (CBA) found in tests that this agent could identify complex network and identity management issues in just 15 minutes that would take a senior engineer hours to diagnose. Furthermore, it can learn from historical events and proactively offer improvement suggestions to prevent future incidents.
AWS emphasizes that these agents not only make teams faster, but also redefine the possibilities of AI collaboration, allowing humans to focus on higher-value innovation and decision-making.
