Agent Id Design Patterns
In brief
The article now uses updated links and clarifies blueprint principals, single- and multi-instance modes, permission inheritance, and delegated versus application access guidance.
What Entra admins need to know
Administrators get clearer guidance for understanding Agent ID deployments and assigning permissions; no configuration change is required.
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.
Key concepts
The following components are the foundation of Microsoft Entra Agent ID. If you're new to them, start with Microsoft Entra Agent ID key conceptsMicrosoft Entra Agent ID key concepts before continuing with this article.
Identity constructs
The following identity constructs are used throughout the patterns described in this article:
Agent identity blueprintAgent identity blueprint: The template and authentication foundation for one or more agent identities. It holds credentials and policies that apply to all agent identities created from it.Agent identity blueprint principalAgent identity blueprint principal: The Microsoft Entra object created when a multi-instance blueprint is added to a tenant. It's what actuallyacquires tokens, creates agent identities, and appears in audit logs on behalf of the blueprint. Single-instance blueprints don't have this principal.Agent identityAgent identity: The runtime identity for a specific AI agent, with its own permissions on downstream resources.Agent's user accountAgent's user account: An optional 1:1 account paired with an agent identity, needed only when the agent must access systems that require a user object.
Permissions models
Blueprint-level permissions and agent identity-level permissions serve different purposes:
BlueprintFleet blueprint permissionsrepresent the minimum permissions shared across all agent identities created from the blueprint. Use inheritable permissions on the blueprint when you want all agent identities to start withcan provide a commonbaseline.baseline through explicitly configured inheritance and consent. Single-instance agents don't have a blueprint principal and receive permissions directly.- Agent identity permissions represent differentiated permissions for a specific agent. Use these permissions when different agents in the same system need different access to downstream resources.
For more information, see Configure inheritable permissions for blueprints and Authorization in Microsoft Entra Agent ID.
Structure: One blueprint → one agent identity
This deployment pattern doesn't by itself select the blueprint's instance mode. A single-instance blueprint and a multi-instance blueprint with one child identity both have one Agent ID, but their client IDs, token acquisition, and permission inheritance differ.
Although a blueprint with a single agent identity might seem redundant, the blueprint gives the agent consistent Conditional Access policies, monitoring, governance, and audit entries — the same infrastructure you need for multi-agent systems, with minimal setup.
Permissions: Grant permissions directly on the agent identity. The blueprint's inheritable permissions aren't typically needed for singleton cases.
Example: An AI sales representative with a real mailbox, listed in the Global Address List, that responds to email and is assigned a human manager in the org chart.
Permissions: Grant the agent's user account the specific Exchange, Teams, and OneDrive permissions it needs. Grant the agent identity application-level permissions for systems that don't require a user object. For more information,delegated access, see Request consent for delegated permissions. For application access, see Permissions and consent in the Microsoft identity platform.
Agent's user account: Required. Create one agent's user account per digital worker agent identity.
Related content
@@ -18,21 +18,21 @@ For step-by-step decision guidance on how many blueprints and agent identities t ## Key concepts -The following components are the foundation of Microsoft Entra Agent ID. If you're new to them, start with [Microsoft Entra Agent ID key concepts](identity-platform/key-concepts.md) before continuing with this article.+The following components are the foundation of Microsoft Entra Agent ID. If you're new to them, start with [Microsoft Entra Agent ID key concepts](key-concepts.md) before continuing with this article. ### Identity constructs The following identity constructs are used throughout the patterns described in this article:-- **[Agent identity blueprint](identity-platform/agent-blueprint.md)**: The template and authentication foundation for one or more agent identities. It holds credentials and policies that apply to all agent identities created from it.-- **[Agent identity blueprint principal](identity-platform/agent-blueprint.md#agent-identity-blueprint-principals)**: The Microsoft Entra object created when a blueprint is added to a tenant. It's what actually acquires tokens, creates agent identities, and appears in audit logs on behalf of the blueprint.-- **[Agent identity](identity-platform/agent-identities.md)**: The runtime identity for a specific AI agent, with its own permissions on downstream resources.-- **[Agent's user account](identity-platform/agent-users.md)**: An optional 1:1 account paired with an agent identity, needed only when the agent must access systems that require a user object.+- **[Agent identity blueprint](agent-blueprint.md)**: The template and authentication foundation for one or more agent identities. It holds credentials and policies that apply to all agent identities created from it.+- **[Agent identity blueprint principal](agent-blueprint.md#agent-identity-blueprint-principals)**: The Microsoft Entra object created when a multi-instance blueprint is added to a tenant. It acquires tokens, creates agent identities, and appears in audit logs on behalf of the blueprint. Single-instance blueprints don't have this principal.+- **[Agent identity](agent-identities.md)**: The runtime identity for a specific AI agent, with its own permissions on downstream resources.+- **[Agent's user account](agent-users.md)**: An optional 1:1 account paired with an agent identity, needed only when the agent must access systems that require a user object. ### Permissions models Blueprint-level permissions and agent identity-level permissions serve different purposes: -- **Blueprint permissions** represent the minimum permissions shared across all agent identities created from the blueprint. Use inheritable permissions on the blueprint when you want all agent identities to start with a common baseline.+- **Fleet blueprint permissions** can provide a common baseline through explicitly configured inheritance and consent. Single-instance agents don't have a blueprint principal and receive permissions directly. - **Agent identity permissions** represent differentiated permissions for a specific agent. Use these permissions when different agents in the same system need different access to downstream resources. For more information, see [Configure inheritable permissions for blueprints](configure-inheritable-permissions-blueprints.md) and [Authorization in Microsoft Entra Agent ID](authorization-agent-id.md).@@ -53,6 +53,8 @@ A single agent that assists with a specific task, commonly built on a low-code o **Structure:** One blueprint → one agent identity +This deployment pattern doesn't by itself select the blueprint's [instance mode](agent-blueprint.md#instance-modes-preview). A single-instance blueprint and a multi-instance blueprint with one child identity both have one Agent ID, but their client IDs, token acquisition, and permission inheritance differ.+ Although a blueprint with a single agent identity might seem redundant, the blueprint gives the agent consistent Conditional Access policies, monitoring, governance, and audit entries — the same infrastructure you need for multi-agent systems, with minimal setup. **Permissions:** Grant permissions directly on the agent identity. The blueprint's inheritable permissions aren't typically needed for singleton cases.@@ -115,7 +117,7 @@ Each digital worker requires its own agent's user account. The 1:1 relationship *Example: An AI sales representative with a real mailbox, listed in the Global Address List, that responds to email and is assigned a human manager in the org chart.* -**Permissions:** Grant the agent's user account the specific Exchange, Teams, and OneDrive permissions it needs. Grant the agent identity application-level permissions for systems that don't require a user object. For more information, see [Grant agent access to Microsoft 365](grant-agent-access-microsoft-365.md).+**Permissions:** Grant the agent's user account the specific Exchange, Teams, and OneDrive permissions it needs. For delegated access, see [Request consent for delegated permissions](grant-agent-access-microsoft-365.md). For application access, see [Permissions and consent in the Microsoft identity platform](/entra/identity-platform/permissions-consent-overview). **Agent's user account:** Required. Create one agent's user account per digital worker agent identity. @@ -135,8 +137,8 @@ Creating one agent identity per meeting, per document, or per ephemeral object a ## Related content -- [Agent identity blueprints](identity-platform/agent-blueprint.md)-- [Agent identities](identity-platform/agent-identities.md)-- [Agent's user accounts](identity-platform/agent-users.md)+- [Agent identity blueprints](agent-blueprint.md)+- [Agent identities](agent-identities.md)+- [Agent's user accounts](agent-users.md) - [Plan your agent identity architecture](how-to-plan-agent-identity-architecture.md) - [Configure inheritable permissions for blueprints](configure-inheritable-permissions-blueprints.md) 