Overview of agent identities in Microsoft Entra
In brief
The documentation now explains credential handling for multi-instance and single-instance Agent IDs, including blueprint impersonation, shared client IDs, PKCE, and permission requirements. It also distinguishes object IDs from application (client) IDs.
What Entra admins need to know
Administrators should use these distinctions when configuring credentials, interpreting identifiers, and granting agent access.
This editorial summary was generated by AI from the documentation changes. Verify important details in the full Microsoft Learn article.
Documentation change
The comparison below shows only the changed extract. Use the full-page view for complete context.
Agent identities in Microsoft Entra Agent ID
An agent identity is a special service principal in Microsoft Entra ID. It representsIts blueprint defines authentication settings and holds any client credentials. In multi-instance mode, the blueprint impersonates its child agent identities. In single-instance mode, the blueprint and Agent ID share an identity thatOAuth client ID, and the Agent ID requests tokens directly.
An agent can request application tokens for autonomous work or delegated tokens for a signed-in user. Public clients don't send a client credential. In either case, permissions must be granted before the agent identity blueprint created and is authorized to impersonate. It doesn't have credentials on its own. The agent identity blueprint can acquire tokens on behalf of the agent identity provided the user or tenant admin consented for the agent identity to the corresponding scopes. Autonomous agents acquire app tokens on behalf of the agent identity. Interactive agents called withaccess a user token acquire user tokens on behalf of the agent identity.resource.
Agent identities can be used to:
Identifier. Each agent identity has an
id(also known as object ID), such asaaaaaaaa-1111-2222-3333-bbbbbbbbbb. Microsoft Entra generates theidand uniquely identifies the account within a Microsoft Entra tenant.Credentials.
Agent identities don't have credentials of their own. They relyCredentials are configured on the blueprint, not the agentidentityidentity. Multi-instance identities use blueprintto acquire tokens on their behalf.impersonation; single-instance confidential clients use the blueprint credential directly.Display name. An agent identity's display name is surfaced in many experiences such as the Microsoft Entra admin center, Azure portal, Teams, Outlook, and more. It's the human-friendly name of an agent and can be changed.
Authorizing agent identities
Agent identity is the primary account used by an AI agent to authenticate to various systems. It has unique identifiers such as theIts object ID andidentifies the app ID, which always havedirectory object, while its application (client) ID identifies it in OAuth requests. A single-instance Agent ID shares the same value and can be reliably usedblueprint's client ID but has a separate object ID. Don't substitute one identifier for authentication and authorization decisions.another in Microsoft Graph requests.
Unlike human users, AI agents don't use passwords, Short Message Service (SMS), passkeys, or authenticator apps for authentication. For multi-instance Agent identities don't have credentials of their own. They only authenticate using federated identity credentials (FIC) issued byIDs, the agent identity blueprint. The blueprint holds credentialsauthenticates and provides an impersonation token that itthe Agent ID uses to acquire tokens on behalf of agent identities.as a client assertion. Credentials do not reside on the blueprint, not the agent identity. These credentialCredential types on the blueprint include:
- Federated identity credentials
- Certificates / cryptographic keys
- Client secrets
A single-instance confidential client authenticates directly with its shared client ID and a credential configured on the blueprint. A public client instead uses authorization code flow with PKCE and no client credential. For the requests and consent requirements, see Authenticate a single-instance agent identity.
Agent identities can only be issued tokens in the Microsoft Entra tenant where they're created. They can't access resources or APIs in other tenants.
@@ -2,14 +2,17 @@ title: Overview of agent identities in Microsoft Entra description: Learn about agent identities in Microsoft Entra ID, specialized identity constructs that enable secure authentication and authorization for AI agents in enterprise environments. ms.topic: concept-article-ms.date: 03/30/2026+ms.date: 11/03/2026 ms.reviewer: dastrock+ai-usage: ai-assisted #customer-intent: As a developer or IT administrator, I want to understand what agent identities are in Microsoft Entra ID, how they differ from application and human identities, and why my organization needs them so that I can deploy AI agents securely in our Microsoft Entra environment. --- # Agent identities in Microsoft Entra Agent ID -An agent identity is a special service principal in Microsoft Entra ID. It represents an identity that the agent identity blueprint created and is authorized to impersonate. It doesn't have credentials on its own. The agent identity blueprint can acquire tokens on behalf of the agent identity provided the user or tenant admin consented for the agent identity to the corresponding scopes. Autonomous agents acquire app tokens on behalf of the agent identity. Interactive agents called with a user token acquire user tokens on behalf of the agent identity.+An agent identity is a special service principal in Microsoft Entra ID. Its blueprint defines authentication settings and holds any client credentials. In multi-instance mode, the blueprint impersonates its child agent identities. In single-instance mode, the blueprint and Agent ID share an OAuth client ID, and the Agent ID requests tokens directly.++An agent can request application tokens for autonomous work or delegated tokens for a signed-in user. Public clients don't send a client credential. In either case, permissions must be granted before the agent can access a resource. Agent identities can be used to: @@ -25,7 +28,7 @@ An account used by an AI agent is referred to as an **agent identity**. Much lik - **Identifier**. Each agent identity has an `id` (also known as object ID), such as `aaaaaaaa-1111-2222-3333-bbbbbbbbbb`. Microsoft Entra generates the `id` and uniquely identifies the account within a Microsoft Entra tenant. -- **Credentials**. Agent identities don't have credentials of their own. They rely on the agent identity blueprint to acquire tokens on their behalf.+- **Credentials**. Credentials are configured on the blueprint, not the agent identity. Multi-instance identities use blueprint impersonation; single-instance confidential clients use the blueprint credential directly. - **Display name**. An agent identity's display name is surfaced in many experiences such as the Microsoft Entra admin center, Azure portal, Teams, Outlook, and more. It's the human-friendly name of an agent and can be changed. @@ -39,14 +42,16 @@ These are the basic components of an agent identity that enable secure authentic ## Authorizing agent identities -Agent identity is the primary account used by an AI agent to authenticate to various systems. It has unique identifiers such as the object ID and the app ID, which always have the same value and can be reliably used for authentication and authorization decisions.+Agent identity is the primary account used by an AI agent to authenticate to various systems. Its object ID identifies the directory object, while its application (client) ID identifies it in OAuth requests. A single-instance Agent ID shares the blueprint's client ID but has a separate object ID. Don't substitute one identifier for another in Microsoft Graph requests. -Unlike human users, AI agents don't use passwords, Short Message Service (SMS), passkeys, or authenticator apps for authentication. Agent identities don't have credentials of their own. They only *authenticate* using [federated identity credentials (FIC)](/graph/api/resources/federatedidentitycredentials-overview) issued by the agent identity blueprint. *The blueprint holds credentials* that it uses to acquire tokens on behalf of agent identities. Credentials do *not* reside on the agent identity. These credential types on the blueprint include:+Unlike human users, AI agents don't use passwords, Short Message Service (SMS), passkeys, or authenticator apps for authentication. For multi-instance Agent IDs, the blueprint authenticates and provides an impersonation token that the Agent ID uses as a client assertion. Credentials reside on the blueprint, not the agent identity. Credential types on the blueprint include: - Federated identity credentials - Certificates / cryptographic keys - Client secrets +A single-instance confidential client authenticates directly with its shared client ID and a credential configured on the blueprint. A public client instead uses authorization code flow with PKCE and no client credential. For the requests and consent requirements, see [Authenticate a single-instance agent identity](authenticate-single-instance-agent.md).+ Agent identities can only be issued tokens in the Microsoft Entra tenant where they're created. They can't access resources or APIs in other tenants. > [!NOTE] 