Caret: A conversational assistant for Java development within Eclipse

Generative artificial intelligence has prompted the proposal of LLM-based conversational assistants in software development, many integrated into IDEs. However, these are generally not fully open-source and operate as black-boxes; they offer limited support for configuring pre- and post-validation pipelines for generated code; they lack detailed tracking mechanisms of code provenance; and are limited regarding collaborative AI-assisted development. To address these limitations, we present Caret , a conversational assistant for Java development within Eclipse. Its open-source architecture is extensible with new assistive and validation tasks, which can be shared via an Eclipse p2 repository. Moreover, it supports collaboration by keeping a record of its code contributions and user–assistant interactions, and providing statistics on how users utilise the assistant.