Skip to main content

Python Flask extension for using Azure Active Directory with OAuth to protect applications

Project description

Flask Azure AD OAuth Provider

Python Flask extension for securing apps with Azure Active Directory OAuth


This provider defines an AuthLib Resource Protector to authenticate and authorise users and other applications to access features or resources within a Flask application using the OAuth functionality offered by Azure Active Directory, as part of the Microsoft identity platform.

This provider depends on Azure Active Directory, which acts as a identity provider, to issue OAuth access tokens. These contain various claims including the identity of the user and client application (used for authentication) and any permissions assigned or delegated to the user or application (used for authorisation).

This provider will validate and interpret information in these tokens to restrict access to parts of a Flask app.

Specifically this provider supports these scenarios:

  1. application to application
    • supports authentication and authorisation
    • used to allow a client application access to some functionality or resources provided by another application
    • can be used for non-interactive, machine-to-machine, processes (using the OAuth Client Credentials Grant)
    • optionally, uses the identity of the client application for authentication
    • optionally, uses permissions assigned directly to the client application for authorisation
  2. user to application
    • supports authentication and authorisation
    • used to allow users access to some functionality or resources provided by another application
    • can be used for interactive console (using the Device Authorization Grant) or web application (using the OAuth Authorization Code Grant) processes
    • uses the identity of the user, and optionally, the client application they are using, for authentication
    • optionally, uses permissions assigned to the user, permissions delegated by the user to the client application, and/or permissions assigned directly to the client application for authorisation

Other scenarios may be added in future versions of this provider.

Note: This provider does not support client applications requesting tokens from Azure. See the Microsoft Authentication Library (MSAL) for Python package if you need to do this.


This package can be installed using Pip from PyPi:

$ pip install flask-azure-oauth


This provider provides an AuthLib Resource Protector which can be used as a decorator on Flask routes.

A minimal application would look like this:

from flask import Flask

from flask_azure_oauth import FlaskAzureOauth

app = Flask(__name__)

app.config['AZURE_OAUTH_TENANCY'] = 'xxx'
app.config['AZURE_OAUTH_APPLICATION_ID'] = 'xxx'

auth = FlaskAzureOauth()

def unprotected():
    return 'hello world'

def protected():
    return 'hello authenticated entity'

def protected_with_scope():
    return 'hello authenticated and authorised entity'

@auth('required-scope1 required-scope2')
def protected_with_multiple_scopes():
    return 'hello authenticated and authorised entity'

To restrict a route to any valid user or client application (authentication):

  • add the resource protector as a decorator (auth in this example) - for example the /protected route

To restrict a route to specific users (authorisation):

  • add any required Scopes to the decorator - for example the /projected-with-* routes

Independently of these options, it's possible to require specific, trusted, client applications, regardless of the user using them. This is useful in circumstances where a user may be authorised but the client can't be trusted:

  • set the AZURE_OAUTH_CLIENT_APPLICATION_IDS config option to a list of Azure application identifiers

For example:


Configuration options

The resource protector requires two configuration options to validate tokens correctly. These are read from the Flask config object through the init_app() method.

Configuration Option Data Type Required Description
AZURE_OAUTH_TENANCY Str Yes ID of the Azure AD tenancy all applications and users are registered within
AZURE_OAUTH_APPLICATION_ID Str Yes ID of the Azure AD application registration for the application being protected
AZURE_OAUTH_CLIENT_APPLICATION_IDS List[Str] No ID(s) of the Azure AD application registration(s) for the application(s) granted access to the application being protected

Note: If the AZURE_OAUTH_CLIENT_APPLICATION_IDS option is not set, all client applications will be trusted and the azp claim, if present, is ignored.

Before these options can be set you will need to:

  1. register the application to be protected
  2. define the permissions and roles this application supports
  3. register the application(s) that will use the protected application
  4. assign permissions to users and/or client application(s)

Flask session support

This provider extends the AuthLib ResourceProtector to support detecting access tokens stored in the Flask session.

This is intended for browser based applications where the Authorization header cannot be easily set to include the access token. This support will be enabled automatically if an access_token session key is set.

Access token versions

Since version 0.5.0, this provider is compatible with Azure access token versions 1.0 and 2.0. Prior to version 0.5.0 only version 2.0 tokens could be used. See Microsoft's documentation for the differences between token versions.

Note: If you use version 1.0 tokens, this provider expects at least one of the identifierUris property values to be api://{protected_application_id}, where {protected_application_id} is the application ID of the app registration representing the application being protected by this provider. Without this, you will receive errors for an invalid audience.

Applications, users, groups and tenancies

Azure Active Directory has a number of different concepts for agents that represent things being protected and things that want to interact with protected things:

  • applications - represent services that offer, or wish to use, functionality that should be restricted:
    • services offering functionality are protected applications, e.g. an API
    • services wishing to use functionality interactively or non-interactively, are client applications:
      • interactive client applications include self-service portals for example
      • non-interactive client applications include nightly synchronisation tasks for example
  • users - represent individuals that wish to use functionality offered by protected applications, through one or more client applications (e.g. a user may use a self-service portal to access information)
  • groups - represent multiple users, for ease of managing permissions to similar users (e.g. administrative users)

For management purposes, all agents are scoped to an Azure tenancy (with the exception of users that can be used across tenancies).

In the Azure management portal:

  • applications are represented by Application registrations
  • users are represented by users, or optionally groups of users

Permissions, roles and scopes

Azure Active Directory has a number of mechanisms for controlling how agents can interact with each other:

  • roles - functions, designations or labels conferred on users and/or groups (e.g. admins, staff)
  • direct permissions - capabilities of a protected application client applications can use themselves or without the consent of the current user (e.g. machine-to-machine access to, or modification of, data from all users)
  • delegated permissions - capabilities of a protected application the current user allows a client application to use (e.g. interactive access to, or modification of, their data)

Generally, and in terms of the OAuth ecosystem, all of these can be considered as scopes. As discussed in the Usage section, scopes can be used to control who and/or what can use features within protected applications.

Scopes are included the access token generated by a client application (possibly interactively by a user) and presented to the projected application as a bearer token. Azure encodes different mechanisms in different claims:

  • roles - for roles assigned to users and permissions directly assigned to client applications
  • scp - for permissions delegated by the user to a client application

For ease of use, this extension abstracts these two claims into a single set of scopes that can be required for a given route. Multiple scopes can be required (as a logical AND) to allow scopes to be used more flexibly.

Defining permissions and roles within an application

Permissions and roles are defined in the application manifest of each application being protected. They can then be assigned to users, groups and client applications.

  1. register the application to be protected
  2. add permissions to application manifest

For example:

"appRoles": [
    "allowedMemberTypes": [
    "displayName": "List all Foo resources",
    "id": "112b3a76-2dd0-4d09-9976-9f94b2ed965d",
    "isEnabled": true,
    "description": "Allows access to basic information for all Foo resources",
    "value": "Foo.List.All"

Assigning permissions and roles to users and applications

Permissions and roles (collectively, application roles) are assigned through the Azure portal:

  1. define roles and permissions in the protected application
  2. register the client application(s)
  3. assign:

For assigning permissions:

  • permissions can be delegated to client applications, with the agreement of the current user
  • permissions can be directly assigned to client applications, with the agreement of a tenancy administrator

Note: Direct assignment is needed for non-interactive applications, such as daemons.

Registering an application in Azure

Follow these instructions.

Note: These instructions apply both to applications that protected by this provider (protected applications), and those that will be granted access to use such applications, possibly by a user (client applications).

Testing support

For testing applications, a local/test JSON Web Key Set (JWKS) can be used to sign local/test JSON Web Tokens (JWTs) without relying on Azure. Local tokens can include, or not include, arbitrary scopes/roles, which can ensure requirements for specific scopes are properly enforced by this provider.

This requires using local tokens signed by the test keys, and patching the FlaskAzureOauth._get_jwks method to validate tokens using the same test keys.

For example:

import unittest

from http import HTTPStatus
from unittest.mock import patch

from flask_azure_oauth import FlaskAzureOauth
from flask_azure_oauth.mocks.keys import TestJwk
from flask_azure_oauth.mocks.tokens import TestJwt

from examples import create_app

class AppTestCase(unittest.TestCase):
    def setUp(self):
        self.test_jwks = TestJwk()

        with patch.object(FlaskAzureOauth, "_get_jwks") as mocked_get_jwks:
            mocked_get_jwks.return_value = self.test_jwks.jwks()

            # `` should be set to a Flask application, either by direct import, or by calling an app factory
   = create_app()

  ["TEST_JWKS"] = self.test_jwks
            self.app_context =
            self.client =

    def test_protected_route_with_multiple_scopes_authorised(self):
        # Generate token with required roles
        token = TestJwt(
  , roles=["BAS.MAGIC.ADD.Records.Publish.All", "BAS.MAGIC.ADD.Records.ReadWrite.All"]

        # Make request to protected route with token
        response = self.client.get(
            "/protected-with-multiple-scopes", headers={"authorization": f"bearer { token.dumps() }"}
        self.assertEqual(HTTPStatus.OK, response.status_code)

    def test_protected_route_with_multiple_scopes_unauthorised(self):
        # Generate token with no scopes
        token = TestJwt(

        # Make request to protected route with token
        response = self.client.get(
            "/protected-with-multiple-scopes", headers={"authorization": f"bearer { token.dumps() }"}
        self.assertEqual(HTTPStatus.FORBIDDEN, response.status_code)


This project is developed as a Python library. A bundled Flask application is used to simulate its usage and to act as framework for running tests etc.

$ git clone
$ cd flask-azure-oauth

Development environment

Docker and Docker Compose are required to setup a local development environment of this application.

If you have access to the BAS GitLab instance, you can pull the application Docker image from the BAS Docker Registry. Otherwise you will need to build the Docker image locally.

# If you have access to
$ docker login
$ docker-compose pull
# If you don't have access:
$ docker-compose build

Code Style

PEP-8 style and formatting guidelines must be used for this project, with the exception of the 80 character line limit.

Black is used to ensure compliance, configured in pyproject.toml.

Black can be integrated with a range of editors, such as PyCharm, to perform formatting automatically.

To apply formatting manually:

$ docker-compose run app black flask_azure_oauth/

To check compliance manually:

$ docker-compose run app black --check flask_azure_oauth/

Checks are ran automatically in Continuous Integration.


Python dependencies for this project are managed with Poetry in pyproject.toml.

Non-code files, such as static files, can also be included in the Python package using the include key in pyproject.toml.

Adding new dependencies

To add a new (development) dependency:

$ docker-compose run app ash
$ poetry add [dependency] (--dev)

Then rebuild the development container, and if you can, push to GitLab:

$ docker-compose build app
$ docker-compose push app

Updating dependencies

To update dependencies:

$ docker-compose run app ash
$ poetry update

Then rebuild the development container, and if you can, push to GitLab:

$ docker-compose build app
$ docker-compose push app

Supported Python versions

This project is only tested against the Python version used in the project container.

Other Python versions may be compatible with this project but these are not tested or officially supported.

A minimum Python version is set in pyproject.toml.

Static security scanning

To ensure the security of this API, source code is checked against Bandit for issues such as not sanitising user inputs or using weak cryptography.

Warning: Bandit is a static analysis tool and can't check for issues that are only be detectable when running the application. As with all security tools, Bandit is an aid for spotting common mistakes, not a guarantee of secure code.

To check manually from the command line:

$ docker-compose run app bandit -r .

Checks are ran automatically in Continuous Integration.


Integration tests

This project uses integration tests to ensure features work as expected and to guard against regressions and vulnerabilities.

The Python UnitTest library is used for running tests using Flask's test framework. Test cases are defined in files within tests/ and are automatically loaded when using the test Flask CLI command included in the local Flask application in the development environment.

To run tests manually:

$ docker-compose run -e FLASK_ENV=testing app flask test --test-runner text

To run tests manually using PyCharm, use the included App (Tests) run/debug configuration.

Tests are ran automatically in Continuous Integration.

Continuous Integration

All commits will trigger a Continuous Integration process using GitLab's CI/CD platform, configured in .gitlab-ci.yml.

Test/Example applications

For verifying this provider works for real-world use-cases, a test Flask application is included in examples/ This test application acts as both an application providing access to, and accessing, protected resources. It can use a number of application registrations registered in the BAS Web & Applications Test Azure AD.

These applications allow testing different versions of access tokens for example. These applications are intended for testing only. They do not represent real applications, or contain any sensitive or protected information.

To test requesting resources from protected resources as an API, set the appropriate config options and run the application container:

$ docker-compose run app
$ flask access-resource [resource]

To test requesting resources from protected resources as a browser application, set the appropriate config options and start the application container:

$ docker-compose up

Terraform is used to provision the application registrations used:

$ cd provisioning/terraform
$ docker-compose run terraform
$ az login --allow-no-subscriptions
$ terraform init
$ terraform validate
$ terraform apply

Note: Several properties in the application registration resources require setting once the registration has been initially made (identifiers for example). These will need commenting out before use.

Some properties, such as client secrets, can only be set once applications have been registered in the Azure Portal.

Terraform state information is held in the BAS Terraform Remote State project (internal).


Python package

This project is distributed as a Python package, hosted in PyPi.

Source and binary packages are built and published automatically using Poetry in Continuous Delivery.

Package versions are determined automatically using the support/python-packaging/ script.

Continuous Deployment

A Continuous Deployment process using GitLab's CI/CD platform is configured in .gitlab-ci.yml.

Release procedure

For all releases:

  1. create a release branch
  2. close release in
  3. push changes, merge the release branch into master and tag with version

The project will be built and published to PyPi automatically through Continuous Deployment.


The maintainer of this project is the BAS Web & Applications Team, they can be contacted at:

Issue tracking

This project uses issue tracking, see the Issue tracker for more information.

Note: Read & write access to this issue tracker is restricted. Contact the project maintainer to request access.


© UK Research and Innovation (UKRI), 2019 - 2020, British Antarctic Survey.

You may use and re-use this software and associated documentation files free of charge in any format or medium, under the terms of the Open Government Licence v3.0.

You may obtain a copy of the Open Government Licence at

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for flask-azure-oauth, version 0.5.0
Filename, size File type Python version Upload date Hashes
Filename, size flask_azure_oauth-0.5.0-py3-none-any.whl (23.8 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size flask-azure-oauth-0.5.0.tar.gz (28.9 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page