Generated tests, run against reality.
A test infrastructure where generated tests exercise the same execution path the production estate would. Modern test runners on the outside, real COBOL execution on the inside. The combination is the point: familiar tooling, genuine behaviour, no fiction in the middle.
How it works
Familiar on the surface, real underneath.
Real execution
Modern runners, real runtime
Tests are written for modern test runners on the outside but execute real COBOL on the inside. The runner is familiar; the thing under test is genuine.
Read more Validated end to end
Docker-validated five-phase harnesses
Client-specific harnesses run a five-phase cycle — covering flows such as commission, premium, policy maintenance, duplicate-client merge, reinstatement and sum-assured — each validated in Docker.
Read more Produced autonomously
Built by the Epic Loop
These harnesses are now produced by the autonomous Epic Loop: a deterministic controller drives an AI maker to build the full suite, at graded autonomy, with every decision logged.
Read more The rule that makes it worth anything
No simulation.
A test that runs against a mock is measuring the mock. The value of behavioural testing on a legacy estate comes entirely from running against the real thing, under real data, so that a passing test is evidence and not decoration.
See it on your code
Generate a test suite for one of your programs.
Bring a program and its expected behaviour. We will generate a suite, run it against the real runtime, and show you where reality and expectation diverge.