Amazon announces Kiro, a new long running autonomous intelligent software coding agent. Kiro was unveiled in July. It expands the company’s current AI coding assistant, which was primarily meant to produce operational code for live deployments.
What makes Kiro really unique, though, is its ability to maintain “persistence context across sessions.” This incredible feature was largely achieved thanks to an engineering process we call “spec-driven development.” This process gives Kiro a chance to learn the finer details of active projects by watching team members work through their workflows, and by reading existing code. As a result, Kiro can be assigned complex tasks and work autonomously for hours or even days, requiring minimal human supervision.
The AI agent can independently address multi-step, complex tasks. For instance, it can debug all 15 layers of a complex corporate software stack with a single prompt. It works by gathering human feedback during the entire coding process, so developers can guide, approve, or even set Kiro straight on any assumptions being made. This back and forth process ensures that Kiro comes up with the most exact specifications to bring to life.
Kiro has potential to be much more productive than coding in the classic way. Matt Garman, CEO of AWS, emphasized this potential, stating, “You simply assign a complex task from the backlog and it independently figures out how to get that work done.” It’s a powerful capability that developers love. They tend to want to limit the scope and test results very quickly before proceeding.
Even with the promise of AI coding agents such as Kiro, the industry still has issues to work through. Modern multimodal large language models (LLMs), such as OpenAI’s GPT‑5.1-Codex-Max, are plagued by hallucination and factual inaccuracy problems. As a result, developers frequently are left to fulfill the responsibilities of “babysitters.” Although GPT‑5.1-Codex-Max is intended for long-winded runs—24 hours, maximum—developers continue to find themselves hovering over their outputs with heightened vigilance.
Kiro’s development is another big step forward for Amazon in artificial intelligence and software development. Amazon incorporates spec-driven development into its coding processes. This strategy accelerates the development cycle and lightens the load for human developers. As enterprises continue to lean more heavily on automation, Kiro’s features have the potential to flip the script on creating and managing software.
