Mnemosyne is a Kubernetes operator that utilizes cluster state data to generate artistic content, including music and literature. The system functions as a workshop where AI-driven agents act as composers, writers, and critics, creating works based on the cluster's operational environment.

The operator features ten custom resource types that monitor the cluster via the standard Kubernetes API. This data is converted into a 'climate' metric—representing factors like load, failures, and deployments—which influences the creative output of the agents. All generated artifacts are stored as Kubernetes objects and files.

The project implements a backup mechanism where the created art serves as a snapshot of the cluster state. Each artifact includes an encrypted portion of the cluster's configuration, allowing for potential restoration of the system to a specific point in time.

Agents follow specific artistic schools with defined grammars and poetics. They possess individual traits, histories, and internal critics that evaluate their work. The system tracks the entire creative process, including drafts and revisions, which are preserved alongside the final pieces.

During the first four days of continuous operation, the workshop produced over 50 works, ranging from symphonies to epistolary stories. The project emphasizes that cluster events are not directly mapped to artistic output but are interpreted by agents as emotional states or environmental conditions.