Great question! In Prefect 3.x, “background tasks” let you run a Prefect task asynchronously without running a flow and without blocking your app. You submit work with

task.delay(...)

, and a Prefect Task Worker picks it up and executes it in the background. Why it’s useful - Don’t block your app: Return HTTP responses immediately while work runs in the background. - Fire-and-forget: Perfect for emails, notifications, webhooks, ETL kicks, etc. - Decouple and scale: Run workers separately from your app and scale them independently. - Simpler than a full queue stack: Acts like a lightweight alternative to Celery/RQ for many use cases. How it works (quick example) 1) Define a task

from prefect import task

@task
def process_order(order_id: str):
    # do work...
    return f"processed {order_id}"

2) Run a Task Worker (in a separate process/container)

from prefect.task_worker import serve

if __name__ == "__main__":
    serve(process_order)  # worker will execute background tasks for this function

3) Submit background work from anywhere (web app, script, another service)

# returns immediately with a PrefectDistributedFuture
fut = process_order.delay("ORD-123")

# optionally, wait for the result later (in the same or different process)
result = fut.result()  # blocks until done

Key points - API: Use

my_task.delay(...)

(alias of

apply_async(...)

). There’s no

prefect.background_task

context manager. - Return type:

delay()

gives a PrefectDistributedFuture you can poll or resolve later. - Execution: Requires a running Task Worker (

prefect.task_worker.serve(...)

) connected to your Prefect server/Cloud. - When to use: Web endpoints, event handlers, async pipelines, and any “queue this work and move on” pattern. Docs and examples - Concepts: Background tasks (concepts) - How-to: Run background tasks - Example app: Background tasks example app If you can share your use case (e.g., FastAPI endpoint, emails, data processing), I can tailor a minimal example to match it.