Open-source MLOps · Apache-2.0
The unified layer forML and AI.
Orchestrate training pipelines and production AI agents on the tools, clouds, and environments you already use — without rewriting your stack.
ZenML is the open-source platform for production ML systems, built by ZenML GmbH and used by ML and AI teams worldwide. Modular, infrastructure-agnostic, no lock-in.
one pipeline · any stack
Proof, not promises
Stack-agnostic by design
Bring your own stack.
ZenML doesn’t replace your tools — it connects them. Choose the pieces you already run and keep the same pipeline code as your stack evolves.
Orchestrator
Cloud
Experiment tracker
orchestrator: Kubeflow
cloud: AWS
tracker: MLflow
Switch any layer later — your pipeline code stays the same. 60+ integrations and counting.
Browse all integrationsWhat the platform gives you
Production ML, without rewriting your stack.
Portable, reproducible pipelines
Write a pipeline once and run it on any orchestrator and cloud. ZenML tracks every step, artifact and model so a run from last quarter rebuilds exactly today.
- Pipelines and stacks across any cloud
- Model registry, lineage and reproducibility
- Smart caching and deduplication of steps
60+ integrations, no lock-in
Plug into the orchestrators, clouds, trackers and model registries you already use — swap any of them without touching pipeline code.
Governance, security and access
Role-based access, secrets and audit trails through ZenML Pro keep teams and regulated workloads in line.
Built for agents too
Checkpoint, recover from crashes and replay agent runs — bring the same rigor you expect from training to production AI agents.
Two products · one team
Start open source. Scale when you’re ready.
ZenML meets teams where they are: a free, self-hostable framework to begin, a managed control plane when you grow, and replay tooling for production agents.
The open-source framework for portable, production-ready ML pipelines. Self-host the server and keep full control of your stack.
- Pipelines, stacks and artifact tracking
- Run the server on your own infrastructure
A managed control plane for teams: collaboration, pipeline monitoring, role-based access and governance on top of the open-source core.
- Team collaboration and RBAC
- Pipeline monitoring and governance
Replay your agent’s real runs with one thing changed, and see what would have happened — debugging and evaluation built for production agents.
- Checkpoint and replay real runs
- Crash recovery for agent flows
Questions, answered
Before you
book a demo.
Straight answers on licensing, lock-in and what runs where. Still curious? The docs and Slack go deeper.
Read the docsIs ZenML really open source?
Yes. The core framework is licensed under Apache-2.0 and the full server is self-hostable — see the repository at github.com/zenml-io/zenml. ZenML Pro is an optional managed control plane on top of the same open-source core.
Do I have to replace my current MLOps tools?
No. ZenML is stack-agnostic: it connects to the orchestrators, clouds, experiment trackers and model registries you already run. There are 60+ integrations including Kubeflow, Airflow, Vertex AI, MLflow and Weights & Biases.
Which clouds and orchestrators are supported?
ZenML runs on AWS, Google Cloud and Azure as well as hybrid and on-prem setups, and orchestrates with Kubeflow, Airflow, Kubernetes, Vertex AI and more. You can switch any layer later without rewriting your pipeline code.
What is the difference between ZenML and ZenML Pro?
ZenML open source is the free, self-hosted framework. ZenML Pro adds a managed control plane — team collaboration, pipeline monitoring, role-based access and governance. Pricing is published at zenml.io/pricing.
Does ZenML work for AI agents, not just training pipelines?
Yes. Alongside training pipelines, ZenML supports production AI agents with checkpointing, crash recovery and replay through Kitaru, so you can debug and evaluate agent runs with the same reproducibility you expect from ML.
How do I get started or talk to the team?
Read the docs at docs.zenml.io, star and clone the repo on GitHub, join the community Slack, or book a demo to see ZenML Pro with your own stack in mind.
Get started
Put your ML and AI
on one layer.
Start with the open-source framework today, or book a demo to see ZenML Pro mapped onto the stack you already run.