Handle errors and exceptions in MSAL for Python

In MSAL for Python, most errors are conveyed as a return value from the API call. The error is represented as a dictionary containing the JSON response from the Microsoft identity platform.

  • A successful response contains the "access_token" key. The format of the response is defined by the OAuth2 protocol. For more information, see 5.1 Successful Response
  • An error response contains "error" and usually "error_description". The format of the response is defined by the OAuth2 protocol. For more information, see 5.2 Error Response

When an error is returned, the "error" key contains a machine-readable code. If the "error" is, for example, an "interaction_required", you may prompt the user to provide additional information to complete the authentication process. If the "error" is "invalid_grant", you may prompt the user to reenter their credentials. The following snippet is an example of error handling in MSAL for Python.

from msal import ConfidentialClientApplication

authority_url = "https://login.microsoftonline.com/your_tenant_id"
client_id = "your_client_id"
client_secret = "your_client_secret"
scopes = ["https://graph.microsoft.com/.default"]

app = ConfidentialClientApplication(client_id, authority=authority_url, client_credential=client_secret)

result = app.acquire_token_silent(scopes=scopes, account=None)

if not result:
    result = app.acquire_token_silent(scopes=scopes)

if "access_token" in result:
    print("Access token: %s" % result["access_token"])
    print("Error: %s" % result.get("error"))

When an error is returned, the "error_description" key also contains a human-readable message, and there is typically also an "error_code" key which contains a machine-readable Microsoft identity platform error code. For more information about the various Microsoft identity platform error codes, see Authentication and authorization error codes.

In MSAL for Python, exceptions are rare because most errors are handled by returning an error value. The ValueError exception is only thrown when there's an issue with how you're attempting to use the library, such as when API parameter(s) are malformed.

Conditional Access and claims challenges

When getting tokens silently, your application may receive errors when a Conditional Access claims challenge such as MFA policy is required by an API you're trying to access.

The pattern for handling this error is to interactively acquire a token using MSAL. This prompts the user and gives them the opportunity to satisfy the required Conditional Access policy.

In certain cases when calling an API requiring Conditional Access, you can receive a claims challenge in the error from the API. For instance if the Conditional Access policy is to have a managed device (Intune) the error will be something like AADSTS53000: Your device is required to be managed to access this resource or something similar. In this case, you can pass the claims in the acquire token call so that the user is prompted to satisfy the appropriate policy.

Retrying after errors and exceptions

MSAL makes HTTP calls to the Azure AD service, and occasionally failures can occur. For example the network can go down or the server is overloaded.

MSAL Python 1.11+ automatically performs one retry attempt for you. You may customize this behavior by following the http_client customization instructions.

HTTP 429

When the Service Token Server (STS) is overloaded with too many requests, it returns HTTP error 429 with a hint about how long until you can try again in the Retry-After response field.

Your app was expected to throttle the subsequent requests, and only retry after the specified period.

MSAL Python 1.16+ makes it easy for you to retry an authentication request on-demand (for example, whenever the end-user clicks the sign-in button again), MSAL Python 1.16+ would automatically throttle those retry attempts by returning same error response from an HTTP cache, and only sending out a real HTTP call when that call is attempted after the specified period.

By default, this throttle mechanism works by saving throttle information into a built-in in-memory HTTP cache. You may provide your own dict-like object as the HTTP cache, which you can control how to persist its content. See MSAL Python API documentation for more details.

Next steps