What can Kiro do that Copilot can't? Amazon launches its own AI IDE

Contents

Amazon has unveiled Kiro, a standalone AI-powered integrated development environment (IDE) that goes beyond traditional code autocompletion and focuses on specification-driven software development. Developed by a small, independent team within AWS, Kiro is primarily designed to help medium and large enterprises implement complex software projects in a structured and efficient manner.

Kiro is technically based on a fork of Code OSS, the open-source core of Visual Studio Code, but is designed as a standalone IDE. According to AWS developer advocate Nathan Peck, Kiro is deliberately “lightly decoupled from AWS” to appeal to developers outside the AWS ecosystem. It can be used without an AWS account, for example via Google or GitHub.

Structure through specifications

At the heart of Kiro are so-called specs, structured requirements, design, and task descriptions organized in Markdown files (requirements.md, design.md, tasks.md). The requirements follow the EARS standard (Easy Approach to Requirements Syntax), which was originally developed by Rolls Royce. These specifications serve as a reference framework for both the AI and the developers throughout the entire project.

Agentic hooks and security modes

Another central element are hooks, automated actions triggered by events such as file saving or creation. They generate tests, perform security checks, or update documentation. Hooks are described in so-called Agentic Actions and are configurable. Kiro works with Claude Sonnet 4.0 as its standard LLM.

A practical report at devclass.com shows that special caution is required in autopilot mode, where the AI is allowed to modify the code independently. Alternatively, supervised mode can be activated, in which users must approve all changes. Critical actions (e.g., installation of npm modules) always require explicit consent. Approval lists can be defined for greater control.

Comparison with GitHub Copilot and other tools

While GitHub Copilot is based on general code suggestions and primarily helps individual developers or small teams, Kiro offers a much more comprehensive range of functions for companies. It is familiar with internal APIs, deployment patterns, review comments, and architectural decisions.

In practice, this is evident in code reviews, compliance with security guidelines (e.g., SOC2, HIPAA), and the standardization of deployment configurations. Kiro can even help with the onboarding of new team members by documenting and making internal knowledge accessible.

Open architecture and pricing model

Kiro is currently in the preview phase and can be tested free of charge. After that, a tiered pricing model is planned: the free version includes 50 “agentic interactions” per month and user, Pro accounts cost $19/month (1,000 interactions), Pro+ $39/month (3,000 interactions). Additional interactions cost $0.04 each. An interaction can take several minutes, for example, if Kiro is processing a complex task with multiple iterations.

Amazon's development environment particularly supports the languages TypeScript/JavaScript, Python, and Java. There are restrictions for .NET developers, as Kiro relies on the Open VSX registry and does not support proprietary extensions from Microsoft (e.g., C#).

Structure instead of chaos in AI coding

Kiro clearly sets itself apart in an increasingly crowded market for AI development tools with its specification orientation and governance features. It aims not only to enable faster coding but also to promote structured, maintainable, and organization-compliant software development. While human developers will remain indispensable, Kiro could take over many routine tasks and help teams move from prototype to production with less friction – across platforms.

Further information can be found in the announcement post on the official website.

(mdo)