Dagster is a cloud-native data orchestrator for the development, production, and observation of data assets (tables, datasets, ML models, reports). It exposes a declarative programming model (assets/jobs defined as Python functions) and provides integrated lineage, observability, and best-in-class testability. Dagster emphasizes a strong developer workflow—enabling local iterative development (e.g., dagster dev) with high parity to production deployments on Docker, Kubernetes, or Dagster Cloud—and a pluggable config and resource system to point to different external systems for testing and production. For sensitive environments it supports a hybrid architecture where the control plane can run in Dagster Cloud while orchestration code and data access remain in the customer environment.