ZenML

license

Apache-2.0(OSI)

source

public(zenml-io/zenml is the open-source server and SDK)

core-gated

gated(ZenML Pro ships as separate container images privately hosted in ZenML's GCP Artifact Registry and AWS ECR repositories, with access granted on request

Apache-2.0 source, with functionality withheld from it. ZenML's own in-repository documentation states the split: the open-source edition has no role-based access control and no workspaces or projects, while Pro adds those plus the Model Control Plane, the Artifact Control Plane and run snapshots. What makes this gating rather than a separate product sold alongside is the packaging - Pro's container images are privately hosted in ZenML's own GCP Artifact Registry and AWS ECR repositories with access granted on request, and a self-hosted Pro deployment pulls them with an image-pull secret. That is a closed component of the deployment you run, the langgraph shape, rather than the langchain shape where a hosted platform sits beside a complete open core.

• https://github.com/zenml-io/zenml recorded 2026-09-15

  Repository page for zenml-io/zenml: Apache-2.0 license, public and unarchived. Establishes the license and that the source is public.

• https://raw.githubusercontent.com/zenml-io/zenml/main/docs/book/getting-started/zenml-pro/README.md recorded 2026-09-15

  ZenML's own in-repository OSS versus Pro feature comparison: OSS marked 'No RBAC' and 'No workspaces/projects', Pro adding the Model Control Plane, Artifact Control Plane and Run Snapshots, and offering SaaS, hybrid or self-hosted deployment.

• https://raw.githubusercontent.com/zenml-io/zenml/main/docs/book/getting-started/zenml-pro/deploy-prerequisites.md recorded 2026-09-15

  ZenML Pro's deployment prerequisites: 'Container images are privately hosted in the ZenML GCP Artifact Registry and ZenML AWS ECR repositories. Access to these registries is granted upon request', with a contact address. Establishes that the withheld functionality ships as a closed package rather than as a separate hosted product.

The PyPI download API reports 125,127 downloads of zenml in the trailing 30 days, inside the 100K-1M band for level 3. PyPI is the product's declared distribution channel.

• https://pypistats.org/api/packages/zenml/recent recorded 2026-09-15

  Trailing-window download counts for the declared PyPI artifact: 125,127 downloads in the trailing month.

Band 3, and it was drafted at 1. The floor band says the product schedules nothing and hands the whole pipeline to an orchestrator it does not own, and ZenML does not quite fit that: the local and Docker orchestrators are its own code, and they decide when each step runs and dispatch it, which is band 3's requirement. Its remote flavors - Airflow, Kubeflow, SageMaker, AzureML, Databricks, Tekton, SkyPilot - do delegate wholly, and that is the product's point. Banding it at the floor required a distinction the ladder does not make, between scheduling onto a local machine and scheduling onto a remote one, so the honest placement is level with the other products whose scheduler dispatches work and lets something else choose the machine.

• https://docs.zenml.io/stacks/stack-components/orchestrators recorded 2026-09-15

  ZenML's orchestrator documentation: the orchestrator is 'an essential component in any MLOps stack as it is responsible for running your machine learning pipelines' and 'a mandatory component in the ZenML stack', with flavors listed for local, Docker, Kubernetes, Airflow, Kubeflow, SageMaker, AzureML, Databricks, Tekton, SkyPilot VM and others.