Microsoft Entra Agent ID

Create and manage OAuth2-capable identities for AI agents using Microsoft Graph beta API.

Preview API — All Agent Identity endpoints are under /beta only. Not available in /v1.0.

Before You Start

Search microsoft-docs MCP for the latest Agent ID documentation:

• Query: "Microsoft Entra agent identity setup"
• Verify: API parameters match current preview behavior

Conceptual Model

Agent Identity Blueprint (application)        ← one per agent type/project
  └── BlueprintPrincipal (service principal)   ← MUST be created explicitly
        ├── Agent Identity (SP): agent-1       ← one per agent instance
        ├── Agent Identity (SP): agent-2
        └── Agent Identity (SP): agent-3

Prerequisites

PowerShell (recommended for interactive setup)

# Requires PowerShell 7+
Install-Module Microsoft.Graph.Beta.Applications -Scope CurrentUser -Force

Python (for programmatic provisioning)

pip install azure-identity requests

Required Entra Roles

One of: Agent Identity Developer, Agent Identity Administrator, or Application Administrator.

Environment Variables

AZURE_TENANT_ID=<your-tenant-id>
AZURE_CLIENT_ID=<app-registration-client-id>
AZURE_CLIENT_SECRET=<app-registration-secret>

Authentication

⚠️ DefaultAzureCredential is NOT supported. Azure CLI tokens contain Directory.AccessAsUser.All, which Agent Identity APIs explicitly reject (403). You MUST use a dedicated app registration with client_credentials flow or connect via Connect-MgGraph with explicit delegated scopes.

PowerShell (delegated permissions)

Connect-MgGraph -Scopes @(
    "AgentIdentityBlueprint.Create",
    "AgentIdentityBlueprint.ReadWrite.All",
    "AgentIdentityBlueprintPrincipal.Create",
    "User.Read"
)
Set-MgRequestContext -ApiVersion beta

$currentUser = (Get-MgContext).Account
$userId = (Get-MgUser -UserId $currentUser).Id

Python (application permissions)

import os
import requests
from azure.identity import ClientSecretCredential

credential = ClientSecretCredential(
    tenant_id=os.environ["AZURE_TENANT_ID"],
    client_id=os.environ["AZURE_CLIENT_ID"],
    client_secret=os.environ["AZURE_CLIENT_SECRET"],
)
token = credential.get_token("https://graph.microsoft.com/.default")

GRAPH = "https://graph.microsoft.com/beta"
headers = {
    "Authorization": f"Bearer {token.token}",
    "Content-Type": "application/json",
    "OData-Version": "4.0",  # Required for all Agent Identity API calls
}

Core Workflow

Step 1: Create Agent Identity Blueprint

Sponsors are required and must be User objects — ServicePrincipals and Groups are rejected.

import subprocess

# Get sponsor user ID (client_credentials has no user context, so use az CLI)
result = subprocess.run(
    ["az", "ad", "signed-in-user", "show", "--query", "id", "-o", "tsv"],
    capture_output=True, text=True, check=True,
)
user_id = result.stdout.strip()

blueprint_body = {
    "@odata.type": "Microsoft.Graph.AgentIdentityBlueprint",
    "displayName": "My Agent Blueprint",
    "sponsors@odata.bind": [
        f"https://graph.microsoft.com/beta/users/{user_id}"
    ],
}
resp = requests.post(f"{GRAPH}/applications", headers=headers, json=blueprint_body)
resp.raise_for_status()

blueprint = resp.json()
app_id = blueprint["appId"]
blueprint_obj_id = blueprint["id"]

Step 2: Create BlueprintPrincipal

This step is mandatory. Creating a Blueprint does NOT auto-create its service principal. Without this, Agent Identity creation fails with: 400: The Agent Blueprint Principal for the Agent Blueprint does not exist.

sp_body = {
    "@odata.type": "Microsoft.Graph.AgentIdentityBlueprintPrincipal",
    "appId": app_id,
}
resp = requests.post(f"{GRAPH}/servicePrincipals", headers=headers, json=sp_body)
resp.raise_for_status()

If implementing idempotent scripts, check for and create the BlueprintPrincipal even when the Blueprint already exists (a previous run may have created the Blueprint but crashed before creating the SP).

Step 3: Create Agent Identities

agent_body = {
    "@odata.type": "Microsoft.Graph.AgentIdentity",
    "displayName": "my-agent-instance-1",
    "agentIdentityBlueprintId": app_id,
    "sponsors@odata.bind": [
        f"https://graph.microsoft.com/beta/users/{user_id}"
    ],
}
resp = requests.post(f"{GRAPH}/servicePrincipals", headers=headers, json=agent_body)
resp.raise_for_status()
agent = resp.json()

API Reference

All endpoints use base URL: https://graph.microsoft.com/beta

Required Permissions

There are 18 Agent Identity-specific Graph application permissions. Discover all:

az ad sp show --id 00000003-0000-0000-c000-000000000000 \
  --query "appRoles[?contains(value, 'AgentIdentity')].{id:id, value:value}" -o json

Grant admin consent (required for application permissions):

az ad app permission admin-consent --id <client-id>

Admin consent may fail with 404 if the service principal hasn't replicated. Retry with 10–40s backoff.

Cleanup

# Delete Agent Identity
requests.delete(f"{GRAPH}/servicePrincipals/{agent['id']}", headers=headers)

# Delete BlueprintPrincipal (get SP ID first)
sps = requests.get(
    f"{GRAPH}/servicePrincipals?$filter=appId eq '{app_id}'",
    headers=headers,
).json()
for sp in sps.get("value", []):
    requests.delete(f"{GRAPH}/servicePrincipals/{sp['id']}", headers=headers)

# Delete Blueprint
requests.delete(f"{GRAPH}/applications/{blueprint_obj_id}", headers=headers)

Best Practices

1. Always create BlueprintPrincipal after Blueprint — not auto-created; implement idempotent checks on both
2. Use User objects as sponsors — ServicePrincipals and Groups are rejected
3. Handle permission propagation delays — after admin consent, wait 30–120s; retry with backoff on 403
4. Include OData-Version: 4.0 header on every Graph request
5. Use Workload Identity Federation for production auth — for local dev, use a client secret on the Blueprint (see references/oauth2-token-flow.md)
6. Set identifierUris on Blueprint before using OAuth2 scoping (api://{app-id})
7. Never use Azure CLI tokens for API calls — they contain Directory.AccessAsUser.All which is hard-rejected
8. Check for existing resources before creating — implement idempotent provisioning

References

External Links