How to Use XCom with TaskFlow API in Airflow

In Airflow's TaskFlow API, use the @task decorator to create tasks that return values, which are automatically pushed to XCom. To pull data from another task, call the task function with .output to get the XCom value, enabling easy data sharing between tasks.

Syntax

The TaskFlow API uses Python functions decorated with @task to define tasks. When a task function returns a value, Airflow automatically pushes it to XCom. To use the value in another task, call the first task function and access its .output attribute.

Key parts:

• @task: marks a Python function as an Airflow task.
• Return value: automatically stored in XCom.
• .output: used to pull the XCom value from a task.

python

from airflow.decorators import dag, task
from airflow.utils.dates import days_ago

@dag(start_date=days_ago(1), schedule_interval='@daily')
def example_dag():

    @task
    def push_data():
        return 'hello from XCom'

    @task
    def pull_data(data):
        print(f'Received: {data}')

    data = push_data()
    pull_data(data.output)

example_dag_instance = example_dag()

Example

This example shows two tasks: one pushes a string to XCom by returning it, and the other pulls it by accessing .output. The second task prints the received value.

python

from airflow.decorators import dag, task
from airflow.utils.dates import days_ago

@dag(start_date=days_ago(1), schedule_interval='@daily', catchup=False)
def xcom_taskflow_example():

    @task
    def push_message():
        return 'Data passed via XCom'

    @task
    def pull_message(message):
        print(f'Pulled message: {message}')

    msg = push_message()
    pull_message(msg.output)

example_dag = xcom_taskflow_example()

Output

Pulled message: Data passed via XCom

Common Pitfalls

• Not returning a value from the task function means nothing is pushed to XCom.
• Trying to access XCom values without using .output on the task instance will fail.
• Using ti.xcom_pull() manually is unnecessary with TaskFlow API and can complicate code.

python

from airflow.decorators import dag, task
from airflow.utils.dates import days_ago

@dag(start_date=days_ago(1), schedule_interval='@daily', catchup=False)
def wrong_xcom_usage():

    @task
    def no_return():
        print('No return, no XCom')

    @task
    def wrong_pull():
        # This will fail because no_return() returns None
        print(no_return())  # Wrong: prints None, no XCom used

    no_return()
    wrong_pull()

example_dag = wrong_xcom_usage()

Quick Reference

Concept	Usage	Notes
@task decorator	Marks a function as a task	Required for TaskFlow API tasks
Return value	Returned value is pushed to XCom	Must return to share data
.output	Access XCom value from a task	Use on task function call
ti.xcom_pull()	Legacy way to pull XCom	Avoid with TaskFlow API
Task dependencies	Set by calling tasks in order	Ensures correct execution sequence

Key Takeaways

Use @task decorator to create tasks that automatically push return values to XCom.

Access XCom data by calling the task function and using .output in downstream tasks.

Always return data from tasks to share it via XCom with TaskFlow API.

Avoid manual ti.xcom_pull() calls when using TaskFlow API for cleaner code.

Define task order by calling tasks in sequence to manage dependencies.