Write Python. Ship Python.

The workflow orchestration framework for Python. From script to scale in a single decorator.

9.6M+ downloads/mo

22.1k+ stars

405+ contributors

New release 6 days ago

Try Prefect Cloud Self-host instead

Trusted in Production

One automation platform
to rule them all.

One engine for Data, ML, and Agents. Unify your entire stack on a single control plane.

Agents

Turn fragile LLM loops into resilient systems with human-in-the-loop control.

Data Science & ML

Scale training from local experiments to thousands of parallel runs.

Data Engineering

Make revenue-impacting decisions on reliable dbt and SQL pipelines.

Durable Background Tasks

Handle emails, user-facing jobs, and async processing without a worker fleet.

Platform Engineering

Unify hundreds of microservices on a single self-serve control plane.

Python orchestration

Just add a decorator.

Run your existing Python code as-is. No rigid DAG structures, no custom DSLs—Prefect fits how you already work.

See how it works

etl_pipeline.py

@flow(retries=3)

def process_data():

data = extract()

transform(data)

load(data)

Status--:--

Waiting

State Tracking

Auto-retries

Artifacts

Durable Execution

Don't replay. Resume.

Recover from failures instantly without re-running expensive work. Prefect persists results to guarantee exactly-once execution for any Python code—no rewrite required.

Learn about durable execution

Workflow Recovery

Run 1

extract()

transform(data)

load(data)

Retry 1

extract()

transform(data)

load(data)

Retry 2

extract()

transform(data)

load(data)

Observed by default

Debug with a map, not a flashlight

Flat logs miss the big picture. See exactly where tasks failed in the timeline and access the relevant logs instantly—no grepping required.

Explore observability

running

striped-pelican7 task runs

get-directory

install-deps

clone-repo

build-project

Run Logs

Initializing run environment...

Event-driven Automation

React to anything

Trigger workflows from webhooks, cloud events, or state changes. Prefect's active event bus detects what happens (or doesn't happen) in real-time, so you can stop polling and start acting.

Explore automations

Work Pools

Portable by default

Work pools decouple your code from where it runs. Switch from Docker to Kubernetes to serverless—your workflows don't change.

Learn about work pools

Docker

Kubernetes

ECS

Google Cloud Run

Azure ACI

Modal

Docker

Kubernetes

ECS

Google Cloud Run

Azure ACI

Modal

Docker

Kubernetes

ECS

Google Cloud Run

Azure ACI

Modal

pipeline.py

1@flow

def pipeline():

3data = extract()

4transform(data)

5load(data)

No changes

Choose how you run Prefect

Self-host with control or go fully managed in the cloud

Open source

Prefect Core

Self-hosted orchestration with full VPC control

Learn more

Recommended

Prefect Cloud

Managed orchestration with SSO, autoscale, and zero maintenance

Learn more

Why teams switch to Prefect

Keep Python. Cut orchestration overhead. Ship faster.

See how Snorkel AI did it

Customer Story

How Snorkel AI runs 1,000+ flows per hour on Prefect OSS

Snorkel AIcase-study

We improved throughput by 20x with Prefect. It's our workhorse for asynchronous processing—a Swiss Army knife. We run about a thousand flows an hour and we're perfectly fine since most of these are network bound.

Smit Shah

Director of Engineering

20x

Throughput improvement

1,000+

Flows per hour

10,000s

Daily workflow executions

Self-hosted on Kubernetes

Incremental migration

Eliminated custom infrastructure

Read the full case study →

Fully Permissive

Apache 2.0. No surprises.

Fork it. Modify it. Build commercial products on top. The Prefect engine is Apache 2.0—the freedom your team needs, forever.

Read full license

Apache License

Version 2.0, January 2004

Permissions:

✓ Commercial use

✓ Modification

✓ Distribution

✓ Patent use

✓ Private use

The fastest way to deploy Prefect

Prefect Cloud is a managed, high-availability deployment of Prefect OSS. Same developer experience, turnkey data platform.

Try Prefect Cloud Free Compare Cloud vs OSS

Trusted in Production

One automation platform
to rule them all.

One engine for Data, ML, and Agents. Unify your entire stack on a single control plane.

Agents

Turn fragile LLM loops into resilient systems with human-in-the-loop control.

Data Science & ML

Scale training from local experiments to thousands of parallel runs.

Data Engineering

Make revenue-impacting decisions on reliable dbt and SQL pipelines.

Durable Background Tasks

Handle emails, user-facing jobs, and async processing without a worker fleet.

Platform Engineering

Unify hundreds of microservices on a single self-serve control plane.

Python orchestration

Just add a decorator.

Run your existing Python code as-is. No rigid DAG structures, no custom DSLs—Prefect fits how you already work.

See how it works

etl_pipeline.py

@flow(retries=3)

def process_data():

data = extract()

transform(data)

load(data)

Status--:--

Waiting

State Tracking

Auto-retries

Artifacts

Durable Execution

Don't replay. Resume.

Recover from failures instantly without re-running expensive work. Prefect persists results to guarantee exactly-once execution for any Python code—no rewrite required.

Learn about durable execution

Workflow Recovery

Run 1

extract()

transform(data)

load(data)

Retry 1

extract()

transform(data)

load(data)

Retry 2

extract()

transform(data)

load(data)

Observed by default

Debug with a map, not a flashlight

Flat logs miss the big picture. See exactly where tasks failed in the timeline and access the relevant logs instantly—no grepping required.

Explore observability

running

striped-pelican7 task runs

get-directory

install-deps

clone-repo

build-project

Run Logs

Initializing run environment...

Event-driven Automation

React to anything

Trigger workflows from webhooks, cloud events, or state changes. Prefect's active event bus detects what happens (or doesn't happen) in real-time, so you can stop polling and start acting.

Explore automations

Work Pools

Portable by default

Work pools decouple your code from where it runs. Switch from Docker to Kubernetes to serverless—your workflows don't change.

Learn about work pools

Docker

Kubernetes

ECS

Google Cloud Run

Azure ACI

Modal

Docker

Kubernetes

ECS

Google Cloud Run

Azure ACI

Modal

Docker

Kubernetes

ECS

Google Cloud Run

Azure ACI

Modal

pipeline.py

1@flow

def pipeline():

3data = extract()

4transform(data)

5load(data)

No changes

Choose how you run Prefect

Self-host with control or go fully managed in the cloud

Open source

Prefect Core

Self-hosted orchestration with full VPC control

Learn more

Recommended

Prefect Cloud

Managed orchestration with SSO, autoscale, and zero maintenance

Learn more

Why teams switch to Prefect

Keep Python. Cut orchestration overhead. Ship faster.

See how Snorkel AI did it

Customer Story

How Snorkel AI runs 1,000+ flows per hour on Prefect OSS

Snorkel AIcase-study

Smit Shah

Director of Engineering

20x

Throughput improvement

1,000+

Flows per hour

10,000s

Daily workflow executions

Self-hosted on Kubernetes

Incremental migration

Eliminated custom infrastructure

Read the full case study →

Fully Permissive

Apache 2.0. No surprises.

Fork it. Modify it. Build commercial products on top. The Prefect engine is Apache 2.0—the freedom your team needs, forever.

Read full license

Apache License

Version 2.0, January 2004

Permissions:

✓ Commercial use

✓ Modification

✓ Distribution

✓ Patent use

✓ Private use

The fastest way to deploy Prefect

Prefect Cloud is a managed, high-availability deployment of Prefect OSS. Same developer experience, turnkey data platform.

Try Prefect Cloud Free Compare Cloud vs OSS

Write Python. Ship Python.

The workflow orchestration framework for Python. From script to scale in a single decorator.

One automation platform to rule them all.

Agents

Data Science & ML

Data Engineering

Durable Background Tasks

Platform Engineering

Just add a decorator.

Don't replay. Resume.

Debug with a map, not a flashlight

React to anything

Portable by default

Choose how you run Prefect

Prefect Core

Prefect Cloud

Why teams switch to Prefect

How Snorkel AI runs 1,000+ flows per hour on Prefect OSS

Apache 2.0. No surprises.

The fastest way to deploy Prefect

Write Python. Ship Python.

The workflow orchestration framework for Python. From script to scale in a single decorator.

One automation platform to rule them all.

Agents

Data Science & ML

Data Engineering

Durable Background Tasks

Platform Engineering

Just add a decorator.

Don't replay. Resume.

Debug with a map, not a flashlight

React to anything

Portable by default

Choose how you run Prefect

Prefect Core

Prefect Cloud

Why teams switch to Prefect

How Snorkel AI runs 1,000+ flows per hour on Prefect OSS

Apache 2.0. No surprises.

The fastest way to deploy Prefect

One automation platform
to rule them all.

One automation platform
to rule them all.