An agent-oriented software-testing system uses Large Language Models to drive the workflow end to end with minimal human intervention: LLM agents generate unit tests, visualize the call graph of the code under test, and execute the tests while producing structured reports. Evaluated on multiple Python and Java applications, the system attains high structural code coverage and operates efficiently, illustrating how LLM-powered agents can streamline the generation-to-reporting pipeline while surfacing scalability and accuracy limits.
This paper investigates an agent-oriented approach to automated software testing powered by Large Language Models (LLMs), aiming to reduce manual effort while improving test quality. The authors buil…