Netflix/metaflow

What is it?

Key features

• 01Pythonic API for defining multi-step flows that run locally or in the cloud
• 02Scale steps horizontally and vertically on CPUs and GPUs
• 03Built-in experiment tracking, versioning and result visualization
• 04Dependency management for running flows remotely
• 05One-step deployment to production workflow orchestrators with event triggering

Use cases

• Prototyping an ML model in a notebook and later moving it to production
• Running large parallel training or data processing jobs on cloud compute
• Managing versioned models and artifacts across a data science team
• Deploying recurring ML pipelines triggered by events

Quick start

1. 1

   Install Metaflow from PyPI

   pip install metaflow
2. 2

   Alternatively, install from conda-forge

   conda install -c conda-forge metaflow
3. 3

   Follow the official tutorial to create and run your first flow

Things to know

• !Full benefits such as cloud scaling and production orchestration require configuring cloud infrastructure, which takes setup effort.

Topics

FAQ

What is metaflow?

A Python framework for building, scaling and deploying AI and ML workflows, from notebook prototypes to production orchestrators. Metaflow is a Python framework for building and managing AI and ML systems. You write a workflow as a Python flow made of steps, then run it locally for prototyping or scale individual steps out to cloud compute clusters using CPUs or GPUs. It includes built-in experiment tracking, versioning of data and artifacts, and visualization of results.

Is metaflow free to use? What license does it have?

metaflow is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.

Is metaflow still maintained?

Actively maintained. The last commit to metaflow was on Oct 16, 2026.

What can I use metaflow for?

Prototyping an ML model in a notebook and later moving it to production Running large parallel training or data processing jobs on cloud compute Managing versioned models and artifacts across a data science team Deploying recurring ML pipelines triggered by events

How do I install metaflow?

Install Metaflow from PyPI: pip install metaflow Alternatively, install from conda-forge: conda install -c conda-forge metaflow Follow the official tutorial to create and run your first flow

Where can I get metaflow?

The source code and installation instructions are on GitHub: https://github.com/Netflix/metaflow