Forward engineering AWS Transform-generated business requirements with OpenAI Codex

This article describes how AWS Transform and OpenAI Codex work together to modernize mainframe applications through reverse and forward engineering with AI agents.

• AWS Transform reverse-engineers COBOL, PL/I, and JCL code to extract business functions and rules with source-line traceability
• Codex forward-engineers approved specifications into cloud-native microservices, tests, and infrastructure using agentic capabilities
• The Reimagine workflow separates reverse engineering, forward engineering, and deployment phases with human validation at each step
• Codex uses sandboxed execution with build-verify-fix loops to autonomously repair code and infrastructure defects
• Traceability annotations (@LegacyRules, @CoversLegacyRules) maintain evidence chains from COBOL source through deployed code and tests
• Codex plugins package reusable skills and connectors for governance, repeatability, and scaling across teams
• CI/CD integration enables parallel cloud tasks, code review with custom rules, and headless validation through the Codex SDK

The approach combines AI-driven code generation with deterministic build gates and human expertise to accelerate enterprise-scale mainframe modernization while maintaining behavioral equivalence and audit trails.