AWS re:Invent 2025 - Kiro meets SaaS: Generating multi-tenant agentic applications with a GenAI IDE

Generating Multi-Tenant SaaS Applications with a GenAI IDE

Ambitious Goal: Building a Production-Ready SaaS Solution on AWS

• The presenters aimed to build a comprehensive SaaS solution on AWS, including:
  • A multi-tenant control plane for SaaS management operations
  • A demo tenant application for service management, with AI agents
• This was to be built using the SaaS Builder Toolkit and the Well-Architected Framework, without shortcuts

Challenges of Building SaaS Today

• Fundamental dilemma: Shipping features and innovating faster while managing complex architectures
• Need to handle multi-tenancy, ensure security, maintain quality, and avoid technical debt
• Balancing speed and quality is the key challenge for SaaS builders

Leveraging Kiro, the GenAI IDE

• Kiro provides capabilities that enable building complex SaaS solutions:
  • Spec-driven development: Kiro generates comprehensive specs with requirements and architecture
  • Agent hooks: Automated workflows that can run in the background to ensure quality
  • Deep context: Kiro can connect to MCP servers to expand its knowledge

Incorporating the SaaS Well-Architected Lens

• The SaaS Well-Architected Lens addresses unique multi-tenancy challenges
• It provides a common language and principles for architectural decisions
• Key pillars:
  • Operational excellence: Tenant-aware monitoring
  • Security: Absolute tenant isolation
  • Reliability: Blast radius containment
  • Performance efficiency: Tenant-based scaling
  • Cost optimization: Protecting the business's bottom line

Leveraging the SaaS Builder Toolkit

• The SaaS Builder Toolkit is an open-source infrastructure tooling from AWS
• It solves common SaaS problems like tenant onboarding, user management, and isolation
• The toolkit's architecture includes a control plane and an application plane, connected by Amazon EventBridge

Collaborative Development with Kiro

• Kiro and the developer work together in a spec-driven, iterative process:
  • Review specification documents
  • Execute tasks step-by-step, with Kiro generating the code
  • Review results, provide feedback, and address issues
  • Run build and unit tests to verify functionality
• This process resulted in:
  • 7 independently deployable components
  • 226 test files, with a 1:4 test-to-production code ratio
  • 327 documentation files, including API references and user guides
• The entire solution, including a full-fledged SaaS application, was created by one person in 2.5 weeks without writing any code

Challenges and Lessons Learned

• Testing and debugging issues: Ensuring a consistent test framework
• Agent Core and Strands: Educating Kiro about new technologies
• Missing requirements: Iterating on the specification to fill gaps

Key Takeaways

1. Quality is fundamental: Focus on building high-quality specifications and solutions
2. Clarity is the ultimate accelerator: The more clarity in requirements, the better the outcome
3. Partnership, not magic: Kiro is an AI assistant, not a magic button - it requires a collaborative effort

Conclusion: SaaS Isn't Dead, It Just Smells Like GenTech

• The presented solution demonstrates the power of leveraging GenAI tools like Kiro to build production-ready SaaS applications in days, not weeks or months.
• SaaS is alive and thriving, with the integration of cutting-edge AI technologies.