From Mainframes to Microservices: Specification-Driven Mainframe Modernization with AI Agents

This article describes a specification-driven approach to mainframe modernization using AWS Transform and Kiro, which reverse-engineers legacy applications into cloud-native microservices through deterministic extraction, domain-driven design, and automated code generation.

• AWS Transform performs deep code analysis, extracts business rules, and generates technology-agnostic EARS-format requirements with full traceability
• Kiro applies Domain-Driven Design principles to identify bounded contexts and decompose legacy requirements into microservice specifications
• Specification-Driven Development transforms architectural blueprints into implementation requirements, design documents, and executable tasks for code generation
• Property-Based Tests validate generated code against behavioral contracts with randomized inputs, tracing failures back to original COBOL business rules
• Three-level specification pipeline ensures nothing is lost: Legacy Requirements → Microservices Specifications → Modernization Requirements
• Human-In-The-Loop checkpoints enable architects to verify specifications before code generation proceeds

This workflow compresses mainframe modernization timelines by automating requirements extraction, design generation, code synthesis, and validation—replacing years of manual expert work with AI-driven precision and traceability.