Prefect is an orchestration and durable execution platform for workflows written in plain Python. Teams use it to run data pipelines and business-critical jobs reliably in production, with scheduling, retries, state, and recovery built in.

Can my coding agent write Prefect workflows?

Yes. Because a Prefect workflow is ordinary Python with light decorators, a coding agent that already knows Python can generate a working flow, wire up retries, and deploy it. Your engineers review and own the result.

How is Prefect different from Airflow?

Prefect runs the control plane for you, so there is no scheduler to host, upgrade, or babysit. You write workflows as plain Python functions instead of rigid DAG definitions, and durable execution handles retries and recovery without code you maintain.

Do I have to run any infrastructure?

No. Prefect Cloud runs the orchestration control plane. With the hybrid model your code and data stay inside your own environment while scheduling and visibility stay central, which is what clears most security and compliance reviews.

Is Prefect open source?

Yes. The Prefect framework is open source under the Apache 2.0 license. Prefect Cloud adds a managed control plane, enterprise authentication, and higher scale for production teams.

Prefect

Open source48.8k+Log in Sign up

Start with one flow.Scale to millions.

Durable orchestration for data, ML, and agents.

Start free

Book a demowith our team

Looking for the MCP gateway? VisitPrefect Horizon

Trusted in Production by:

Case study →

Enterprise GradeSOC 2 Type IIHIPAAGDPRHybrid & VPC99.99% uptime

One platformfrom ML to AI

Data you can depend on

Replace brittle cron jobs with pipelines that heal themselves, so your data always arrives on time.

from prefect import flow, task
 
@task(retries=3)
def extract(source):
    ...
 
@flow
def daily_revenue_elt():
    load(transform(extract("warehouse")))

Deployments / daily_revenue_elt · k8s-prodRunning · 0:00

Success rate

99.2%▲ +0.3%

Avg duration

1.4s▼ −0.2s

Runs · 24h

2,182▲ +8.4%

p95 latency

4.1s▲ +0.4s

Task timeline6 tasks

ingest3s

normalize6s

validate5s

enrich13s

score5s

load3s

0:000:100:200:30

Teams ship faster on Prefect

From startups to enterprises, teams trust Prefect with the workflows their businesses depend on.

How Ramp moved critical data workflows to Prefect

Watch

How WHOOP cut production incidents by 75%

Watch

How Barstool Sports Scaled Media Data Operations with Prefect

Cash App Gains Flexibility in Machine Learning Workflows with Prefect

How Seven.One Entertainment Orchestrates Data Pipelines with Prefect

Snorkel AI

How Snorkel AI Reliably Executes Thousands of Daily Workflows with Prefect Open Source

How Flatiron Health Orchestrates Cancer Research Data Pipelines

Clearcover

The 5-Year Data Payoff: Why Clearcover Keeps Building on Prefect's Pythonic Foundation

See all customer stories

Annotate your code.

Add @flow to your Python functions for state, logging, and automatic retries. No new framework to learn.

etl_pipeline.py

@flow(retries=3)
def process_data():
    data = extract()
    transform(data)
    load(data)

Our place or yours.

Run on managed compute, keep execution inside your own VPC, or anywhere in between. The same code deploys anywhere.

deploy daily_revenue

Prefect Cloudfully managed

or

Your VPCyour code never leaves

Runs you can trust.

Prefect executes your flows on your infrastructure, retries failures, scales out, and tracks every task, so runs recover on their own.

daily-revenue/running

extract0.3s

transform.map

load

Back to building.

Skip building your own cron-and-Slack monitoring stack. One UI shows every run and flags what's late or failed, so you get your engineering hours back.

integration-tests

Search workspace⌘K

Home

Past 7 days

Flow Runs123,066 total

4,074

9

110,923

1

8,059

Active Work Pools

kubernetes-prd-internal-tools96.26% completed

managed-work-pool100% completed

etl_pipeline.py

@flow(retries=3)
def process_data():
    data = extract()
    transform(data)
    load(data)

deploy daily_revenue

Prefect Cloudfully managed

or

Your VPCyour code never leaves

daily-revenue/running

extract0.3s

transform.map

load

integration-tests

Search workspace⌘K

Home

Past 7 days

Flow Runs123,066 total

4,074

9

110,923

1

8,059

Active Work Pools

kubernetes-prd-internal-tools96.26% completed

managed-work-pool100% completed

Annotate your code.

Add @flow to your Python functions for state, logging, and automatic retries. No new framework to learn.

Never fly blind in production.

See every run across all your teams and infrastructure, and catch problems before they spread.

integration-tests

Search workspace⌘K

Deployments / daily-revenue

Past 7 days

2,182

Runs

<1%

Failure rate

2s

Duration

0s

Lateness

0

SLA violations

110MB

Max memory

positioning-extractionCron ScheduleCompleted

revenue-pipelineCron ScheduleCompleted

hourly-syncCron ScheduleRunning

Open source at the core

Prefect Cloud runs on the open-source Prefect framework. We also build FastMCP, the open-source framework for building MCP servers that connect agents to your tools and data.

Prefect

The orchestration layer for workflows

Turn any Python function into a durable, observable workflow with one decorator.

pipeline.py

from prefect import flow
 
@flow(retries=3)
def daily_report():
    data = extract()
    load(transform(data))

11.8M+/mo

FastMCP

The context layer for agents

Build MCP servers in minutes and connect agents to any system.

server.py

from fastmcp import FastMCP
 
mcp = FastMCP("agent-tools")
 
@mcp.tool
def search(query):
    return db.query(query)

96M+/mo

Questions answered

What is Prefect?: Prefect is an orchestration and durable execution platform for workflows written in plain Python. Teams use it to run data pipelines and business-critical jobs reliably in production, with scheduling, retries, state, and recovery built in.
Can my coding agent write Prefect workflows?: Yes. Because a Prefect workflow is ordinary Python with light decorators, a coding agent that already knows Python can generate a working flow, wire up retries, and deploy it. Your engineers review and own the result.
How is Prefect different from Airflow?: Prefect runs the control plane for you, so there is no scheduler to host, upgrade, or babysit. You write workflows as plain Python functions instead of rigid DAG definitions, and durable execution handles retries and recovery without code you maintain.
Do I have to run any infrastructure?: No. Prefect Cloud runs the orchestration control plane. With the hybrid model your code and data stay inside your own environment while scheduling and visibility stay central, which is what clears most security and compliance reviews.
Is Prefect open source?: Yes. The Prefect framework is open source under the Apache 2.0 license. Prefect Cloud adds a managed control plane, enterprise authentication, and higher scale for production teams.

Your first flow, in minutes.

scheduleretryobservedeploycachetriggerbackfillalert

Start free

Book a demowith our team

Looking for the MCP gateway? VisitPrefect Horizon