You’ve got two good options in Prefect 3.x depending on what you want to test: 1) Fast unit tests (no Prefect backend at all) - Call the underlying Python function with

.fn(...)

— this bypasses Prefect’s engine and will not try to reach the API. - Works for both flows and tasks. Example:

from prefect import flow, task

@task
def add(a, b):
    return a + b

@flow
def my_flow(x, y):
    # for unit tests, call task functions directly too
    return add.fn(x, y)

# no API calls, just pure Python
result = my_flow.fn(1, 2)
print(result)  # 3

Notes: - If your task uses

get_run_logger()

, wrap the call with

disable_run_logger()

in tests:

from prefect.logging import disable_run_logger

with disable_run_logger():
    result = my_task.fn(...)

2) Integration-style tests (exercise Prefect orchestration without an external API) - Use

prefect_test_harness()

to run against a temporary local test database. This lets you call flows normally (e.g.,

.submit()

,

.result()

, states) without connecting to

<http://prefect:4200/api/>

. Example:

from prefect import flow, task
from prefect.testing.utilities import prefect_test_harness

@task
def add(a, b):
    return a + b

@flow
def my_flow(x, y):
    fut = add.submit(x, y)
    return fut.result()

with prefect_test_harness():
    print(my_flow(1, 2))  # 3

Why you’re seeing the error - Calling

my_flow()

(not

.fn

) tells Prefect to create a flow run and contact the API at

<http://prefect:4200/api/>

(from

PREFECT_API_URL

). If that URL isn’t reachable, you’ll get “Cannot create flow run. Failed to reach API…”. - For testing, either: - Use

.fn(...)

to bypass the engine entirely, or - Wrap your tests in

prefect_test_harness()

to run locally without a real backend. If you share what you want to validate (pure function behavior vs. Prefect features like

.submit()

, retries, state), I can suggest the exact pattern to use.