I-edit

Ibahagi sa

NotebookUtils credentials utilities for Fabric

You can use the credentials utilities to get access tokens and manage secrets in Azure Key Vault. The notebookutils.credentials module integrates with Microsoft Entra ID for token acquisition and Azure Key Vault for secret management, so you can connect to Azure resources securely without exposing credentials in code.

The credentials utilities are available in Python, PySpark, Scala, and R notebooks. The examples on this page use Python as the primary language, with Scala and R equivalents shown where the public API supports them.

Important

Never hardcode secrets or credentials directly in notebook code. Always use Azure Key Vault to store sensitive values and retrieve them at runtime with notebookutils.credentials.getSecret.

Constraints and safety

Before you use credentials utilities, be aware of these constraints:

  • Token expiration – Tokens expire after a period. For long-running operations, implement refresh logic to request a new token before expiration.
  • Service principal scope limitations – When running under a service principal, tokens for the pbi audience have restricted scopes compared to user identity.
  • MSAL for full scope – If you need the full Fabric service scope under a service principal, use MSAL authentication instead of getToken.
  • Secret redaction – Notebook outputs automatically redact secret values to prevent accidental exposure.
  • Key Vault permissions – You must have appropriate permissions (Get for reading, Set for writing) on the Azure Key Vault to access or store secrets.
  • Audience changes – Token audience scopes might evolve over time. Verify current scopes in the documentation.

Run the following command to get an overview of the available methods:

notebookutils.credentials.help()

The following table lists the available credentials methods:

Method Signature Description
getToken getToken(audience: String): String Returns a Microsoft Entra token for the specified audience.
getSecret getSecret(akvName: String, secret: String): String Returns the value of a secret from the specified Azure Key Vault.
putSecret putSecret(akvName: String, secretName: String, secretValue: String): String Stores a secret in the specified Azure Key Vault. This method isn't available in the public Scala API.
isValidToken isValidToken(token: String): Boolean Checks whether the given token is valid and not expired. This method isn't available in the public Scala API.

Get token

getToken returns a Microsoft Entra token for a given audience. The following table shows the currently available audience keys:

Audience key Resource Use case
storage Azure Storage Access ADLS Gen2 and Blob Storage
pbi Power BI Call Power BI and Fabric REST APIs
keyvault Azure Key Vault Retrieve Key Vault secrets
kusto Synapse RTA KQL DB Connect to Azure Data Explorer

Run the following command to get the token:

notebookutils.credentials.getToken('audience Key')

Token usage examples

You can use the returned token to authenticate against various Azure services.

Azure Storage

storage_token = notebookutils.credentials.getToken('storage')

Power BI and Fabric REST APIs

import requests

pbi_token = notebookutils.credentials.getToken('pbi')

headers = {
    'Authorization': f'Bearer {pbi_token}',
    'Content-Type': 'application/json'
}

response = requests.get(
    'https://api.powerbi.com/v1.0/myorg/datasets',
    headers=headers
)

if response.status_code == 200:
    datasets = response.json()
    print(f"Found {len(datasets['value'])} datasets")

Azure Data Explorer (Kusto)

kusto_token = notebookutils.credentials.getToken('kusto')

Azure Key Vault

keyvault_token = notebookutils.credentials.getToken('keyvault')

Use tokens with the Azure SDK

Fabric notebooks don't support DefaultAzureCredential directly. You can use a custom credential class as a workaround to pass NotebookUtils tokens to Azure SDK clients.

from azure.core.credentials import AccessToken, TokenCredential
import jwt

class NotebookUtilsCredential(TokenCredential):
    """Custom credential that uses notebookutils tokens for Azure SDK."""

    def __init__(self, audience="storage"):
        self.audience = audience

    def get_token(self, *scopes, claims=None, tenant_id=None, **kwargs):
        token = notebookutils.credentials.getToken(self.audience)

        # Decode token to get expiration time
        token_json = jwt.decode(
            token, algorithms="RS256",
            options={"verify_signature": False}
        )

        return AccessToken(token, int(token_json.get("exp", 0)))

# Example: use with Azure Blob Storage
from azure.storage.blob import BlobServiceClient

account_url = "https://mystorageaccount.blob.core.windows.net"
credential = NotebookUtilsCredential(audience="storage")
blob_client = BlobServiceClient(account_url=account_url, credential=credential)

for container in blob_client.list_containers():
    print(f"Container: {container.name}")

Tip

Tokens expire after a period of time. If your notebook runs long operations, implement refresh logic to request a new token before the current one expires.

Considerations

  • Token scopes with pbi as audience might change over time.

  • When you call notebookutils.credentials.getToken("pbi"), the returned token has limited scope if the notebook runs under a service principal. The token doesn't have the full Fabric service scope. If the notebook runs under the user identity, the token still has the full Fabric service scope, but this might change with security improvements. To ensure that the token has the full Fabric service scope, use MSAL authentication instead of the notebookutils.credentials.getToken API. For more information, see Authenticate with Microsoft Entra ID.

  • The following scopes are available when you call notebookutils.credentials.getToken with the audience key pbi under the service principal identity:

    • Lakehouse.ReadWrite.All – Read and write access to Lakehouse items
    • MLExperiment.ReadWrite.All – Read and write access to Machine Learning Experiment items
    • MLModel.ReadWrite.All – Read and write access to Machine Learning Model items
    • Notebook.ReadWrite.All – Read and write access to Notebook items
    • SparkJobDefinition.ReadWrite.All – Read and write access to Spark Job Definition items
    • Workspace.ReadWrite.All – Read and write access to Workspace items
    • Dataset.ReadWrite.All – Read and write access to Dataset items

Tip

If you need access to additional Fabric services or broader permissions under a service principal, use MSAL for Python to authenticate directly with the full Fabric service scope instead of relying on getToken("pbi").

Get secret

getSecret returns an Azure Key Vault secret for a given Azure Key Vault endpoint and secret name. The call uses your current user credentials to authenticate against Key Vault.

notebookutils.credentials.getSecret('https://<name>.vault.azure.net/', 'secret name')

You can retrieve multiple secrets to build connection strings or configure services:

vault_url = "https://myvault.vault.azure.net/"

db_host = notebookutils.credentials.getSecret(vault_url, "db-host")
db_user = notebookutils.credentials.getSecret(vault_url, "db-user")
db_password = notebookutils.credentials.getSecret(vault_url, "db-password")

connection_string = f"Server={db_host};User={db_user};Password={db_password}"

Note

Notebook outputs automatically redact secret values for security. If you print or display a retrieved secret, the output shows a redacted placeholder instead of the actual value.

Use the fully qualified Key Vault URL in the format https://<vault-name>.vault.azure.net/. You must have appropriate permissions to access the Key Vault and the individual secrets.

Security best practices

Follow these recommendations when you work with credentials in Fabric notebooks:

  • Store all sensitive values in Azure Key Vault. Never embed credentials, connection strings, or API keys directly in notebook code.
  • Don't log secret values. Rely on the automatic secret redaction in notebook outputs. Avoid writing secrets to files or passing them as notebook parameters.
  • Use the correct audience key. Match the audience key to the target Azure resource so the token has only the permissions it needs.
  • Understand the identity context. Know whether your notebook runs under user identity or a service principal, because the available token scopes can differ. Test authentication in both interactive and pipeline contexts.
  • Handle token expiration. Tokens expire. For long-running operations, implement refresh logic to request a new token before the current one expires.
  • Limit Key Vault access. Grant only the minimum required permissions to your Key Vault. Audit secret access through Azure Key Vault diagnostic logs.
  • Use managed identities when possible. Managed identities reduce the need to manage credentials manually and provide a more secure authentication flow.

Put secret

putSecret stores a secret in the specified Azure Key Vault. If the secret already exists, the value is updated.

notebookutils.credentials.putSecret('https://<name>.vault.azure.net/', 'secret name', 'secret value')

You must have appropriate permissions (Set permission) on the Azure Key Vault to write secrets.

vault_url = "https://myvault.vault.azure.net/"

notebookutils.credentials.putSecret(vault_url, "api-key", "my-secret-api-key-value")

Validate token

Use isValidToken to check whether a token is valid and not expired before you call an API with it.

token = notebookutils.credentials.getToken('storage')
is_valid = notebookutils.credentials.isValidToken(token)

if is_valid:
    print("Token is valid")
else:
    print("Token is expired or invalid, requesting a new one")
    token = notebookutils.credentials.getToken('storage')

Related content

  • Last updated on