The world of AI-assisted coding is moving faster than most developers can keep up with. We’ve moved from basic autocompletion to "vibe coding," and now, we are entering the era of Agentic IDEs. At the forefront of this shift is Kiro, a new standalone environment from AWS that promises to bring rigor and engineering discipline to AI-generated code.

What is Kiro?

Kiro is a standalone, AI-native IDE built on the Code OSS foundation (the same open-source core as VS Code). While it looks and feels familiar to any VS Code user, its "brain" is fundamentally different. Kiro is designed to be an autonomous agent that doesn't just suggest code but plans and executes entire features. It is currently the primary platform for what AWS calls Spec-Driven Development (SDD).

The Philosophy: Spec-Driven Development vs. Vibe Coding

Most AI tools today work on a "guess and check" basis—you prompt, it codes, and you fix the errors. AWS calls this "vibe coding." Kiro’s philosophy is the opposite: Think twice, code once.

Kiro enforces a structured workflow where the AI must prove it understands your requirements and the technical architecture before it is allowed to modify a single file. This creates a "North Star" for the AI, preventing it from getting lost in complex refactors or hallucinating incorrect APIs.

How Kiro Works: The Three-Phase Workflow

When you give Kiro a high-level goal, it moves through three version-controlled stages:

  • Requirements Phase: Kiro generates a requirements.md file using EARS (Easy Approach to Requirements Syntax). It lists user stories and acceptance criteria, including edge cases you might have missed.
  • Design Phase: It analyzes your codebase to create a design.md. This includes technical architecture, data flow diagrams (using Mermaid), and specific interface definitions.
  • Task Phase: Finally, it creates a tasks.md checklist. Each task is a discrete step linked back to the original requirements, which the AI then executes one by one.

Key Specifics of the Kiro Environment

  • Agent Hooks: Event-driven triggers that run in the background. For example, a hook can automatically update documentation or run security scans every time you save a file.
  • Agent Steering: You can guide the AI using persistent Markdown files (like tech.md or structure.md) stored in the .kiro/steering/ directory. These act as "laws" for your project’s coding standards.
  • Model Context Protocol (MCP): Kiro can securely connect to external tools, such as live AWS documentation or your team’s internal wiki, to ensure its knowledge is never outdated.

How Kiro Differs from Claude Code

While both are "agentic," they represent two different developer personalities:

  • Architecture: Kiro is a full, separate IDE (a fork of VS Code), while Claude Code is a terminal-native CLI tool.
  • Workflow: Kiro is proactive and formal; it stops you to review plans. Claude Code is reactive and fast; it executes tasks at "thought-speed" directly in your existing terminal.
  • Persistence: Kiro stores its "thinking" in durable Markdown files in your repo. Claude Code's logic is largely ephemeral within the chat session.

The Evolution: Is Kiro the new Amazon Q?

Yes and no. Kiro is the evolution of the agentic experience within the AWS ecosystem.

  • The Amazon Q Developer CLI has officially been updated and rebranded as the Kiro CLI. If you run q update in your terminal, you are effectively migrating to Kiro.
  • However, Amazon Q (the VS Code extension) isn't going away. Amazon Q remains the "generalist assistant" for chat and support, while Kiro is the specialized "lab" for deep, autonomous feature building.

What Amazon Q Users Should Know

If you are already using Amazon Q in VS Code, transitioning to Kiro is straightforward:

  • CLI Transition: Use kiro-cli (or stay with q, which is now aliased to Kiro) for terminal-based agentic work.
  • Settings Migration: Your existing rules, prompts, and MCP configurations from the ~/.aws/amazonq folder are automatically migrated to ~/.kiro during installation.
  • Identity: You can continue using your AWS Builder ID or IAM Identity Center for authentication; your existing subscriptions will carry over.

Conclusion

Kiro is built for developers who are tired of the "AI chaos" of unstructured tools. By making specifications a first-class citizen, it allows dev managers to maintain high quality-bars even when using autonomous agents. Whether you use it as your main IDE or as a specialized tool for complex features, it is a glimpse into the future of production-grade AI engineering.