Elyra extends JupyterLab with an AI-centric toolkit for building and running machine learning workflows. Its centerpiece is a visual pipeline editor: drag notebooks and Python or R scripts onto a canvas, connect them into a DAG, and run the pipeline locally or on Kubeflow Pipelines or Apache Airflow, with dependencies and runtime images handled per node. Elyra has around 2,000 GitHub stars and is in active development, with releases continuing into 2026.
Data science work often stalls between the notebook and production: Elyra bridges that gap inside the environment scientists already use. Beyond the pipeline editor, it adds code-snippet management, a metadata-driven catalog of reusable components, Git integration for notebooks, and hybrid runtime support through Jupyter Enterprise Gateway so heavy steps can execute on remote clusters while you edit locally. Pipelines can be exported and scheduled, and generic plus framework-specific components make it straightforward to standardize team workflows.
Free. Elyra is open source under the Apache-2.0 license and installs from PyPI or conda-forge; pipelines run on your own Jupyter and Kubernetes infrastructure.
Data science and ML teams living in Jupyter who want repeatable, schedulable pipelines without abandoning notebooks for a separate MLOps platform.
Elyra originated in IBM’s data science tooling work and is now a community project under the Elyra AI organisation, with contributions from multiple vendors. It integrates with the broader Jupyter ecosystem rather than replacing it: pipelines can include notebooks from a project, reuse runtime configurations stored as metadata, and be submitted from the command line or a CI job as well as from the visual editor, which makes it practical to promote an exploratory notebook workflow into a scheduled production pipeline.