AWS re:Invent 2025 - Accelerate multi-step SDLC with Kiro (DVT321)

Accelerating Multi-Step SDLC with Kiro: An AI-Powered Approach

Overview

The presentation showcased how AWS is empowering developers to accelerate the software development lifecycle (SDLC) using AI-powered tools, specifically the Kiro platform. The key focus was on leveraging AI agents to assist with each stage of the SDLC, from requirements gathering to deployment, rather than just automating the coding step.

The AIDLC Workflow

• The presenters introduced the AI-Driven Lifecycle (AIDLC) methodology, an open-source framework that guides developers through the end-to-end SDLC using AI agents.
• AIDLC consists of three main phases:
  1. Inception: Covers requirements gathering, system design, and planning.
  2. Construction: Handles the actual building and packaging of the software.
  3. Operation: Addresses deployment and ongoing maintenance.

Configuring the AI Agent

• The presenters demonstrated how to configure a custom AI agent using the Kiro CLI, which involves:
  • Selecting the language model
  • Defining the context window (system prompts, resource files, history, etc.)
  • Integrating external tools via the Model Context Protocol (MCP)
• This custom agent configuration can be shared, versioned, and reused across teams.

Building a "Flappy Kiro" Game

• Using the configured AI agent, the presenters walked through the AIDLC process to build a browser-based game called "Flappy Kiro":
  1. Requirements Gathering: The agent asked clarifying questions to capture the game's functional and non-functional requirements.
  2. System Design: The agent generated Mermaid diagrams to visualize the game's architecture and file structure.
  3. Implementation: The agent automatically generated the HTML, CSS, and JavaScript files needed to implement the game.
  4. Testing: The agent proposed a testing plan, including cross-browser and mobile device testing.

Key Takeaways

• AI can be leveraged throughout the entire SDLC, not just for the coding step.
• Configuring a custom AI agent with the right context and tools is crucial for getting desired results.
• The AIDLC workflow provides a structured approach to incorporating AI into the software development process.
• Kiro offers both a CLI and an IDE, allowing developers to choose their preferred development environment.
• Integrating accessibility (WCAG) requirements upfront is important to ensure compliance.
• Standardizing AI knowledge and agent configurations across an organization can help with adoption and consistency.

Technical Details

• The presenters used the Kiro CLI and a custom agent configuration to build the "Flappy Kiro" game.
• The agent leveraged the Mermaid MCP server to generate architectural diagrams.
• The game was implemented using HTML, CSS, and JavaScript, with the option to integrate AWS services like Amplify and Bedrock.

Business Impact

• The AIDLC workflow and Kiro platform can help organizations accelerate their software development process by automating repetitive tasks and ensuring consistency across projects.
• Incorporating AI throughout the SDLC can lead to faster time-to-market, improved quality, and better alignment with business requirements.
• The ability to configure custom AI agents and share them across teams can foster collaboration and knowledge sharing within an organization.

Examples

• The presenters demonstrated the creation of a "Flappy Kiro" game, where the AI agent handled requirements gathering, system design, and code generation.
• They also discussed the importance of integrating accessibility (WCAG) requirements upfront to ensure compliance, using the game as an example.