Create an AI connection for Azure Foundry (v 3.1.7)

Create an AI connection for Azure AI Foundry in AI Control Tower using the  AI Service Graph Connector for Microsoft (version 3.1.7).

Azure Foundry prerequisites

Complete the following steps in your Azure environment before creating an Azure Foundry connection.

• Configure OAuth Credentials
• The connector uses OAuth to authenticate with Azure APIs. To obtain credentials, register an application in Microsoft Entra ID.

For full instructions, see the Azure documentation

The Azure client application requires the following roles:

• Reader role at the subscription or resource group level to discover resources.
• Azure User role on the Azure AI Foundry resources.

Note:

As of March 2026, ServiceNow supports the New Azure AI Foundry alongside the original Azure AI Foundry. The New Foundry treats each agent version as a distinct entity.

Discovery scope

Configure the scope of Azure Foundry discovery using the following options:

• Tenant-wide discovery (default): Leave the Resource Name and Region fields empty to discover all Al agents across your entire Azure tenant.
• Filter by resource (optional): To limit discovery to specific resources, enter resource names as comma-separated values (examples: resource1, resource2).
• Filter by region (optional): To limit discovery to specific Azure regions, enter region names as comma-separated values (for examples: eastus, westus2).

Before you begin

Role required: sn_ai_disc.discovery_admin and sn_cmdb_int_util.sgc_admin

Account & Resource Hierarchy

The connector supports three Azure service variants, each with its own resource hierarchy:

• ML Services (AI Hub) Subscriptions → Resource Groups → ML Workspaces → Agents
• AI Services/Old Foundry (Cognitive Services) Subscriptions → Resource Groups → Cognitive Services Accounts → Projects → Agents
• New Foundry Subscriptions → Resource Groups → Accounts → Projects → Agents → Agent Versions

The key distinction with New Foundry is that each agent version is treated as a distinct entity, which the other two variants don't support.

Discovered per agent

For each agent discovered across all three variants, the connector collects:

• AI Agents (assistants)- The primary entity.
• AI Models- Deployed models (GPT-4o, Llama, Claude, etc.) via deployment enrichment.
• AI Prompts- System instructions attached to agents.
• AI Tools- With type coverage varying by variant: ML & AI Services: functions, connected_agent, and others.
• Sub-component Relationships- M2M links between agents and their sub agents/tools.
• Usage/Execution Metrics- Aggregated run counts by agent, date, and session.

Procedure

1. Navigate to Al Control Tower workspace > Configurations > AI connections.
2. Select AI connector for Microsoft from the available connectors and then select Create connection.
3. Review setup instructions page displays.
   Note:

   Verify to review the setup instructions and automation script.
4. Select I have read the setup instructions check box.
5. Select Continue.

   Select authentication type page appears.

6. Choose the authentication type from the drop-down and select Submit.
7. Select Client credentials and skip to step 10.
8. Select Certificate-based authentication.
9. Create X.509 certificate:
   Note:

   You can create a new certificate or use an existing one.
   1. Select New.
   2. Enter theName.
   3. Enter the Key store password.
   4. Select +Add file to add an attachment.
   5. Select Save.
   6. Select the newly created certificate and select Continue.
10. Create and test connection:
    1. Under Select source systems select the Azure Foundry check box.
    2. Enter the Connection Name.
    3. Enter the Tenant ID.
    4. Enter the OAuth Client ID.
    5. Enter the Keystore.
    6. Enter the Keystore password.
    7. Enter the Thumbprint.
       Note:

       The Region and Resource name fields are optional. You can enter multiple Environment URLs by separating them with a comma.
11. Configure import schedule:
    1. Open the Azure Foundry scheduled job.
    2. Verify that both the parent-scheduled jobs, Discovery and Execution are active as they’re shipped out inactive.
       Note:

       Ensure to execute the Discovery-scheduled job first.
    3. Select Run according to your preference.
    4. To run frequency by demand, select Execute Now.
       Note:

       This is an optional step as the schedule imports run according to the schedule.
    5. Select Continue.
12. Select Confirm connection setup activity to verify whether the connection was configured.

Result

Select View all connections to review the connection details. The created connection appears in the Installed connections list.