User stories (US) are a key artifact in software engineering inherited from Agile methods that allow functional and non-functional requirements to be expressed from the end user's perspective. However, their study in natural language processing has several limitations. Agile and open source methodologies are partially incompatible: the former assumes close, colocated collaboration, while the latter relies on distributed, asynchronous teams. As a result, there exists few public datasets, and none that combine high-quality US, established metrics, and contextual project elements. In practice, US are often poorly formulated and evaluated through informal and oral reviews such as the “Three Amigos” ceremony. These limitations hinder the progress in US's automatic generation and evaluation. This article presents an Multi- Agents AI tool that generates and evaluates contextualized US, providing both comparable and reproducible results for researchers and practitioners.