Open-Source Script Uses Free AI Tokens to Automate C++ Test Coverage

A developer has published an open-source Python script called MonkeyCode that uses free AI model access to automatically generate C++ unit tests and iteratively improve code coverage. The tool sends a C++ header file to an AI model, compiles and runs the resulting tests, then measures coverage using gcov. Any uncovered lines are fed back to the model as a hint, and the loop repeats until coverage plateaus or hits a target threshold. In a sample run, the script pushed coverage from 42% to 91% across four iterations by progressively catching edge cases like empty strings and null-byte handling. The author cautions that AI-generated tests should not be trusted for safety-critical code, describing the tool as a feedback-driven test generator rather than a replacement for a human test engineer.