Manage agent identities in your organization

Microsoft Entra Agent ID provides a centralized set of tools for managing agent identities across your organization. Agent identities are a distinct identity type in Microsoft Entra ID, designed for AI agents with classification, metadata, and security controls tailored to agentic workloads.

This article covers key agent management tasks: from viewing and disabling agents, to governing access, monitoring activity, and responding to security risks. Whether you're an administrator responsible for tenant-wide agent oversight or a sponsor managing specific agents, this guide provides the information you need to effectively manage agent identities in your organization.

Prerequisites

Different management tasks require different roles and licenses. The following table summarizes the roles you need for each area of agent identity management.

Task Required role Notes
View agent identities Microsoft Entra user account No admin role needed for viewing.
Manage agent identities Agent ID Administrator or Cloud Application Administrator Agent identity owners can manage their own agents without these roles.
Create agent blueprints Agent ID Developer User is added as owner of the blueprint and its service principal.
Configure Conditional Access policies Conditional Access Administrator Requires Microsoft Entra ID P1 license.
Configure Lifecycle Workflows Lifecycle Workflows Administrator  

View agent identities

The Microsoft Entra admin center provides a centralized interface to view all agent identities in your tenant. You can search, filter, sort, and customize columns to find specific agents.

  1. Sign in to the Microsoft Entra admin center.
  2. Browse to Entra ID > Agents > Agent identities.
  3. Select any agent identity to view its details, including name, description, status, owners, sponsors, granted permissions, and sign-in logs.

To search for a specific agent, enter the name or object ID in the search box, or add the Blueprint App ID filter. You can customize which columns are shown by selecting the Choose columns button. Available columns include Name, Created On, Status, Object ID, View Access, Blueprint App ID, Owners and Sponsors, and Uses agent identity.

For detailed instructions on filtering, column customization, and viewing agents from this view, see View and filter agent identities in your tenant.

Manage agent identity blueprints

Agent identity blueprints are the parent definitions from which individual agent identities are created. The admin center lets you view all blueprint principals, manage their permissions, and monitor their activity.

  1. Sign in to the Microsoft Entra admin center.
  2. Browse to Entra ID > Agents > Agent blueprints.
  3. Select any agent identity blueprint principal to manage it.

From a blueprint's management page, you can:

  • View linked agent identities: See all child agent identities created from this blueprint.
  • Manage blueprint access: View, manage, and revoke permissions assigned to the blueprint.
  • Manage owners and sponsors: Owners handle technical administration, while sponsors are accountable for the agent's purpose and lifecycle decisions.
  • View audit logs: Track administrative changes for security and compliance.
  • View sign-in logs: Monitor authentication events.
  • Disable the blueprint: Select Disable in the command bar.

For detailed instructions, see View and manage agent identity blueprints in your tenant.

Configure inheritable permissions for blueprints

Inheritable permissions let agent identities automatically inherit OAuth 2.0 delegated permission scopes from their parent blueprint. When you configure inheritable permissions on a blueprint, newly created agent identities receive a base set of scopes without requiring interactive user or admin consent prompts.

Two inheritance patterns are supported per resource app:

Pattern Description
Enumerated scopes Inherit only explicitly listed scopes. Use for fine-grained control.
All allowed scopes Inherit all available delegated scopes for the resource app. Newly granted scopes on the blueprint are automatically included.

Start with enumerated scopes using only essential permissions, then expand as needed. This approach follows the principle of least privilege and makes it easier to audit which permissions agents actually use.

Key limits:

  • Maximum of 10 resource apps per blueprint.
  • For enumerated scopes, maximum of 40 scopes per resource app.
  • Some high-privilege scopes are blocked by platform policy and can't be inherited.

Inheritable permissions are configured via the inheritablePermissions navigation property on the agentIdentityBlueprint application resource using Microsoft Graph. For step-by-step API examples (add, update, delete), see Configure inheritable permissions for agent identity blueprints.

Control agent access to resources

Conditional Access policies provide tenant-wide controls for agent identity authentication. You can use these policies to block all agent identities or allow only specific agents.

Key points about Conditional Access for agent identities:

  • Policies can scope to All agent identities or All agent users with a Block grant control.
  • Policies can target All resources to prevent agent access across your organization.
  • Policies support Report-only mode for safe evaluation before enforcement.

Important

Conditional Access enforcement applies when an agent identity or agent's user account requests a token for any resource. It does not apply when an agent identity blueprint acquires a token to create agent identities or agent's user accounts.

For detailed policy configuration, step-by-step walkthroughs, and business scenario examples, see Conditional Access for Agent ID.

Monitor agent activity

Sign-in and audit logs

Agent identity activities are captured in Microsoft Entra audit and sign-in logs:

  • Audit logs record agent-related events under the base identity type from which they originate. For example, creating an agent identity user appears as a "Create user" audit activity, and creating an agent identity appears as a "Create service principal."
  • Sign-in logs include the agentSignIn resource type, which provides properties about the agent and its sign-in behavior.

To view agent sign-in logs:

  1. Sign in to the Microsoft Entra admin center as at least a Reports Reader.
  2. Browse to Entra ID > Monitoring & health > Sign-in logs.
  3. Use the following filters:
    • Agent type: Choose from Agent ID user, Agent Identity, Agent Identity Blueprint, or Not Agentic.
    • Is Agent: Choose from No or Yes.

You can also retrieve agent sign-in events using Microsoft Graph:

GET https://microsoftgraph.chinacloudapi.cn/beta/auditLogs/signIns?$filter=signInEventTypes/any(t: t eq 'servicePrincipal') and agent/agentType eq 'AgentIdentity'

For full details on log types and access methods, see Microsoft Entra Agent ID logs.

Govern agent identities and sponsor oversight

Each agent identity should have an accountable human sponsor responsible for lifecycle and access decisions. Key governance behaviors include:

  • Automatic sponsor transfer: If a sponsor leaves the organization, Microsoft Entra ID automatically reassigns sponsorship to the sponsor's manager.
  • Expiration notifications: Sponsors receive notifications when access package assignments approach expiration. Sponsors can request an extension (which triggers a new approval cycle) or let assignments expire.
  • Access request pathways: Access packages can be requested through three pathways: the agent's own programmatic request, a sponsor on behalf of the agent, or an admin direct assignment.

Access packages can grant Security Group memberships, Application OAuth API permissions (including Microsoft Graph application permissions), and Microsoft Entra roles.

For the full governance overview including access package configuration and sponsor policies, see Governing Agent Identities.

Automate sponsor notifications with Lifecycle Workflows

Lifecycle Workflows provides two automated tasks for agent identity sponsorship:

  • Send email to manager about sponsorship changes
  • Send email to cosponsors about sponsor changes

Both tasks are mover and leaver category tasks: they trigger only under mover or leaver workflow templates, not joiner templates. This ensures continuity of sponsorship when an agent's sponsor changes roles or leaves the organization.

For step-by-step workflow configuration, see Agent identity sponsor tasks in Lifecycle Workflows.

Automate agent management at scale

For organizations managing large numbers of agent identities, the following options are available:

  • Multi-select disable: The admin center supports selecting multiple agent identities at once and disabling them in batch from the All agent identities page.
  • Microsoft Graph API: Agent identity endpoints support programmatic management.

Disable or restrict agent identities

Organizations can control agent identity usage at three levels, depending on the scope needed:

Scope What it does Details
Individual agent Disable a specific agent identity to block its access and token issuance. Administrators use the admin center; owners and sponsors use the My Account portal. View and filter agent identities in your tenant · Manage agents in end user experience
Blueprint-level Disable an agent identity blueprint from its management page. This prevents new agent identities from being created from that blueprint and blocks existing ones. View and manage agent identity blueprints in your tenant
Tenant-wide Block all agent identity authentication using Conditional Access policies, and optionally block creation of new agent identities through product-specific controls (Microsoft Entra ID, Microsoft Teams). Disable agent identities in your tenant

Re-enabling a disabled agent identity at any scope restores access and token issuance.

Caution

Globally disabling agent identities can cause existing agents to fail, degrade Microsoft product experiences, and push teams to use less transparent application or service principal identities. Evaluate the impact before enforcing. For a partial approach, use Conditional Access policies to block specific agents rather than all agent identities.