Behavior-first,
test-driven

Mechanical Orchard uses a behavior-first approach: the running legacy system, not its documentation or extracted rules, is the source of truth. Imogen works in three phases.
Analyze reads the source code deterministically to map how the system works and what it depends on.

Rewrite generates modern code and verifies it against real production data, slice by slice.

Orchestrate runs legacy and modern components side by side, so the system stays live throughout the change.

Behavioral equivalence means the modernized code produces the same outputs as the legacy system when given the same inputs, including edge cases, encodings and error handling. Imogen treats it as something to prove, not assume. It captures what the legacy system does through real data flows across component interfaces, then uses that behavioral specification to guide AI code generation and check equivalence at every step.

Imogen follows test-driven development: the tests come first, then the code. To move quickly, Imogen generates tests by running synthetic data through the legacy system’s own code and recording how it behaves. In parallel, Mechanical Orchard instruments the mainframe to capture real production data flows, which serve as the ultimate test. New code is written to pass those tests, transaction by transaction, with outputs compared for byte-level equivalence. Verification is automated, so each component carries evidence of equivalence before it goes live.

Imogen preserves business rules by preserving behavior. It uses the live behavior of the original program as the complete specification, so rules that are undocumented or buried in edge cases are still captured: the new code must reproduce the legacy system’s outputs before it replaces anything. To keep disruption to a minimum, the modern code starts out behaving identically to the original. Once that parity is proven, changing business rules in the new code is the safest way forward, and a faster one.

No. Imogen rewrites and cuts over workloads incrementally, one component at a time. Legacy and modern components run side by side, and each component goes live only when verification evidence shows it behaves like the original. Each component also has a proven fallback, so an issue affects one slice rather than the whole system, and the mainframe stays in service until the last component is replaced.

Imogen exceeds 100,000 lines of verified code per engineer per week. In early 2026 engagements, Imogen moved an extended warranty system for a global manufacturer to AWS in four months, an 80% acceleration over initial estimates, and in the summer of 2026, Imogen completed an 11,578-line job in 28 hours, fully verified. We are seeing continued leaps in speed without sacrificing quality and equivalence.

The software dark factory is Imogen’s autonomous build and verification pipeline, introduced in the Summer 2026 release. It generates the new code and verifies it against the legacy system’s behavior autonomously, with security and compliance checks embedded in the pipeline. The pipeline includes templates specifically tuned to mainframe modernization. Engineers set direction and review evidence rather than hand-writing and hand-testing each component.

Imogen supports COBOL, PL/I, Easytrieve, Assembler and JCL batch workloads; CICS and IMS transaction environments; and Db2, VSAM, ISAM and IDMS data stores, including IDMS ADS. Imogen’s Analyze phase reads this source directly to map dependencies before any code is rewritten.

Imogen rewrites mainframe workloads into modern, maintainable Java or Python that runs on cloud platforms including AWS and Google Cloud. Because the output is modern code backed by a full test suite, changes after cutover are faster and safer to make, whether the customer’s own teams or a system integrator makes them. Once parity is proven, the business can reimagine the application with confidence.

Mechanical Orchard runs Imogen primarily on-premises, in environments the customer controls, and Imogen can work in air-gapped environments. Deployment options are agreed per engagement to meet the customer’s security, data-residency and regulatory requirements.

Imogen and AWS Transform are complementary. AWS Transform’s business rules extraction produces a plain-language map of an application, from lines of business down to component-level rules, which helps teams scope and sequence modernization waves. Since July 2026, those extracted rules flow directly into Imogen, which links each rule to its executable verification and proves the rewritten code matches the legacy system’s behavior on production-representative data.

Imogen integrates with Google’s Mainframe Assessment Tool (MAT), which assesses legacy estates and informs wave planning; the integration was part of Imogen’s Spring 2026 release. Imogen then rewrites and verifies each wave for deployment on Google Cloud, and is available on Google Cloud Marketplace. Imogen also complements Google Cloud Dual Run, which replays live mainframe transactions on a modernized system and compares the results. Dual Run fits after Imogen’s verification: Imogen proves equivalence as each component is built, and Dual Run can add a further check on live production traffic before cutover.

Business rules extraction is a strong starting point for scoping, and it works best alongside behavioral verification. Extraction surfaces a representative subset of a program’s conditional logic, while details such as file organization, character encoding and error handling often stay implicit in the code. Imogen closes that gap by testing the rewritten code against the live behavior of the original system, so rules that extraction misses are still preserved.

System integrators deliver mainframe modernization programs on Imogen and own the client relationship. Mechanical Orchard leads a free two-to-three-week proof of concept on the client’s real code, then trains and certifies the SI’s engineers. From there, the SI carries delivery while Imogen automates analysis, code generation and verification, proving each job equivalent before cutover. For SIs, that means lower delivery risk and cost, whole programs scoped from evidence rather than pilots, the ability to offer outcome-based pricing instead of time and materials, and a faster path to the innovation work clients value most.