Access the Airflow REST API

Cloud Composer 1 | Cloud Composer 2

Apache Airflow has a REST API interface that you can use to perform tasks such as getting information about DAG runs and tasks, updating DAGs, getting Airflow configuration, adding and deleting connections, and listing users.

For an example of using Airflow REST API with Cloud Functions, see Triggering DAGs with Cloud Functions.

Airflow REST API versions

The following Airflow REST API versions are available in Cloud Composer 2:

  • Airflow 2 uses the stable REST API. The experimental REST API is deprecated by Airflow.

  • You can still use the experimental REST API in Airflow 2 if you enable it through an Airflow configuration override, as described further.

Before you begin

Enable the Cloud Composer API.

Enable the API

Enable the stable Airflow REST API

The stable REST API is already enabled by default in Airflow 2.

Cloud Composer uses its own API authentication backend.

Authorization works in the standard way provided by Airflow. When a new user authorizes through the API, the user's account gets the Op role by default.

You can enable or disable the stable REST API, or change the default user role by overriding the following Airflow configuration options:

Section Key Value Notes
api (Airflow 2.2.5 and earlier) auth_backend
(Airflow 2.3.0 and later) auth_backends
airflow.composer.api.backend.composer_auth To disable the stable REST API, change to airflow.api.auth.backend.deny_all
api composer_auth_user_registration_role Op You can specify any other role.

Enable the experimental Airflow REST API

By default, the API authentication feature is disabled in the experimental API. The Airflow web server denies all requests that you make.

To enable the API authentication feature and the Airflow 2 experimental API, override the following Airflow configuration option:

Section Key Value Notes
api (Airflow 2.2.5 and earlier) auth_backend
(Airflow 2.3.0 and later) auth_backends
airflow.api.auth.backend.default The default is airflow.composer.api.backend.composer_auth.
api enable_experimental_api True The default is False.

After you set this configuration option to airflow.api.auth.backend.default, the Airflow web server accepts all API requests without authentication. Even though the Airflow web server itself does not require authentication, it is still protected by Identity-Aware Proxy which provides its own authentication layer.

Allow API calls to Airflow REST API using Webserver Access Control

Depending on the method used to call Airflow REST API, the caller method can be using either IPv4 or IPv6 address. Remember to unblock IP traffic to Airflow REST API using Webserver Access Control.

Use the default configuration option which is All IP addresses have access (default) if you are not sure from which IP addresses your calls to Airflow REST API will be sent.

Make calls to Airflow REST API

This section provides an example Python script which you can use to trigger DAGs with the stable Airflow REST API.

Put the contents of the following example into a file named, and then provide your Airflow UI URL, the name of the DAG, and the DAG run config in the variable values.

from __future__ import annotations

from typing import Any

import google.auth
from google.auth.transport.requests import AuthorizedSession
import requests

# Following GCP best practices, these credentials should be
# constructed at start-up time and used throughout
CREDENTIALS, _ = google.auth.default(scopes=[AUTH_SCOPE])

def make_composer2_web_server_request(
    url: str, method: str = "GET", **kwargs: Any
) -> google.auth.transport.Response:
    Make a request to Cloud Composer 2 environment's web server.
      url: The URL to fetch.
      method: The request method to use ('GET', 'OPTIONS', 'HEAD', 'POST', 'PUT',
        'PATCH', 'DELETE')
      **kwargs: Any of the parameters defined for the request function:
                  If no timeout is provided, it is set to 90 by default.

    authed_session = AuthorizedSession(CREDENTIALS)

    # Set the default timeout, if missing
    if "timeout" not in kwargs:
        kwargs["timeout"] = 90

    return authed_session.request(method, url, **kwargs)

def trigger_dag(web_server_url: str, dag_id: str, data: dict) -> str:
    Make a request to trigger a dag using the stable Airflow 2 REST API.

      web_server_url: The URL of the Airflow 2 web server.
      dag_id: The DAG ID.
      data: Additional configuration parameters for the DAG run (json).

    endpoint = f"api/v1/dags/{dag_id}/dagRuns"
    request_url = f"{web_server_url}/{endpoint}"
    json_data = {"conf": data}

    response = make_composer2_web_server_request(
        request_url, method="POST", json=json_data

    if response.status_code == 403:
        raise requests.HTTPError(
            "You do not have a permission to perform this operation. "
            "Check Airflow RBAC roles for your account."
            f"{response.headers} / {response.text}"
    elif response.status_code != 200:
        return response.text

if __name__ == "__main__":
    # TODO(developer): replace with your values
    dag_id = "your-dag-id"  # Replace with the ID of the DAG that you want to run.
    dag_config = {
        "your-key": "your-value"
    }  # Replace with configuration parameters for the DAG run.
    # Replace web_server_url with the Airflow web server address. To obtain this
    # URL, run the following command for your environment:
    # gcloud composer environments describe example-environment \
    #  --location=your-composer-region \
    #  --format="value(config.airflowUri)"
    web_server_url = (

    response_text = trigger_dag(
        web_server_url=web_server_url, dag_id=dag_id, data=dag_config


For example, the configuration below is not correct

  web_server_url = (

Access Airflow REST API using a service account

The Airflow database limits the length of the email field to 64 characters. Service accounts sometimes have email addresses that are longer than 64 characters. It is not possible to create Airflow users for such service accounts in the usual way. If there is no Airflow user for such a service account, then accessing the Airflow REST API results in HTTP errors 401 and 403.

As a workaround, you can preregister an Airflow user for a service account. To do so, use as the username, and any unique string as the email.

  1. To get NUMERIC_USER_ID for a service account, run:

    gcloud iam service-accounts describe \ \


    • SA_NAME with the service account name.
    • PROJECT_ID with the Project ID.
  2. Create an Airflow user with the Op role for the service account:

    Airflow UI

    1. Go to the Airflow UI.

    2. Go to Security > List Users and click Add a new record. Your Airflow user must have the Admin role to open this page.

    3. Specify as the user name. Replace NUMERIC_USER_ID with the user ID obtained on the previous step.

    4. Specify a unique identifier as the email. You can use any unique string.

    5. Specify the role for the user. For example, Op.

    6. Make sure that Is Active? checkbox is selected.

    7. Specify the first and the last name for the user. You can use any string.

    8. Click Save.


    Run the following Airflow CLI command:

    gcloud composer environments run ENVIRONMENT_NAME \
        --location LOCATION \
        users create -- \
        -u \
        -e UNIQUE_ID  \
        -f UNIQUE_ID \
        -l - -r Op --use-random-password


    • ENVIRONMENT_NAME with the name of the environment.
    • LOCATION with the region where the environment is located.
    • NUMERIC_USER_ID with the user ID obtained in the previous step.
    • UNIQUE_ID with the identifier for the Airflow user. You can use any unique string.
  3. After you create an Airflow user for a service account, a caller authenticated as the service account is recognized as a pre-registered user, and is logged into Airflow.

Scaling Airflow REST API component

Airflow REST API and Airflow UI endpoints are run within the component, i.e. Airflow Webserver. In case, you use REST API intensively, consider increasing CPU and memory parameters to adjust Airflow Webserver resources to the expected load.

What's next