Set up notifications for changes in resource data
The Microsoft Graph API uses a webhook mechanism to deliver change notifications to clients. A client is a web service that configures its own URL to receive change notifications. Client apps use change notifications to update their state upon changes.
After Microsoft Graph accepts the subscription request, it pushes change notifications to the URL specified in the subscription. The app then takes action according to its business logic. For example, it fetches more data, updates its cache and views, and so on.
By default, change notifications do not contain resource data, other than the id. If the app requires resource data, it can make calls to Microsoft Graph APIs to get the full resource. This article uses the user resource as an example for working with change notifications.
An app can also subscribe to change notifications that include resource data, to avoid having to make additional API calls to access the data. Such apps will need to implement extra code to handle the requirements of such notifications, specifically: responding to subscription lifecycle notifications, validating the authenticity of notifications, and decrypting the resource data. For details about how to work with these notifications, see Set up change notifications that include resource data.
Supported resources
An app can subscribe to changes on the Microsoft Graph resources listed in the table, which also indicates the limits that apply for subscriptions to the resources. When any limit is exceeded, attempts to create a subscription will result in an 403 Forbidden error response. The message property of the error response will explain the limit that has been exceeded.
Note
Subscriptions to resources marked with an asterisk (*) are available on the /beta endpoint only.
| Resource | Supported resource paths | Limitations |
|---|---|---|
| Cloud printing printer | Changes when a print job is ready to be downloaded (jobFetchable event): /print/printers/{id}/jobs |
- |
| Cloud printing printTaskDefinition | Changes when there is a valid job in the queue (jobStarted event): /print/printtaskdefinition/{id}/tasks |
- |
| driveItem on OneDrive (personal) | Changes to content within the hierarchy of any folder: /users/{id}/drive/root |
- |
| driveItem on OneDrive for Business | Changes to content within the hierarchy of the root folder: /drives/{id}/root , /users/{id}/drive/root |
- |
| group | Changes to all groups: /groups Changes to a specific group: /groups/{id} Changes to owners of a specific group: /groups/{id}/owners Changes to members of a specific group: /groups/{id}/members |
Maximum subscription quotas: Not supported for Azure AD B2C tenants. A known issue for the subscription changeType. |
| list under a SharePoint site | Changes to content within the list: /sites/{site-id}/lists/{list-id} |
- |
| Microsoft 365 group conversation | Changes to a group's conversations: groups/{id}/conversations |
- |
| Outlook message | Changes to all messages in a user's mailbox: /users/{id}/messages , /me/messages Changes to messages in a user's Inbox: /users/{id}/mailFolders('inbox')/messages , /me/mailFolders('inbox')/messages |
A maximum of 1,000 active subscriptions per mailbox for all applications is allowed. |
| Outlook event | Changes to all events in a user's mailbox: /users/{id}/events , /me/events |
A maximum of 1,000 active subscriptions per mailbox for all applications is allowed. |
| Outlook personal contact | Changes to all personal contacts in a user's mailbox: /users/{id}/contacts , /me/contacts |
A maximum of 1,000 active subscriptions per mailbox for all applications is allowed. |
| Security alert | Changes to a specific alert: /security/alerts/{id} Changes to filtered alerts: /security/alerts/?$filter={parameters} |
- |
| Teams callRecord | Changes to all call records: /communications/callRecords |
Maximum subscription quotas: |
| Teams chat | Changes to any chat in the tenant: /chats Changes to a specific chat: /chats/{id} |
Maximum subscription quotas: |
| Teams chatMessage | Changes to chat messages in all channels in all teams: /teams/getAllMessages Changes to chat messages in a specific channel: /teams/{id}/channels/{id}/messages Changes to chat messages in all chats: /chats/getAllMessages Changes to chat messages in a specific chat: /chats/{id}/messages Changes to chat messages in all chats a particular user is part of: /users/{id}/chats/getAllMessages |
Maximum subscription quotas: |
| Teams channel | Changes to channels in all teams: /teams/getAllChannels Changes to channel in a specific team: /teams/{id}/channels |
Maximum subscription quotas: |
| Teams conversationMember | Changes to membership in a specific team: /teams/{id}/members Changes to membership in all channels under a specific team: teams/{id}/channels/getAllMembers Changes to membership in a specific chat: /chats/{id}/members Changes to membership in all chats: /teams/getAllMembers |
Maximum subscription quotas: |
| Teams onlineMeeting * | Changes to an online meeting: /communications/onlineMeetings/?$filter=JoinWebUrl eq {joinWebUrl} |
|
| Teams presence * | Changes to a single user's presence: /communications/presences/{id} Changes to multiple user presences: /communications/presences?$filter=id in ({id},{id}...) |
|
| Teams team | Changes to any team in the tenant: /teams Changes to a specific team: /teams/{id} |
Maximum subscription quotas: |
| todoTask | Changes to all task in a specific task list: /me/todo/lists/{todoTaskListId}/tasks |
- |
| user | Changes to all users: /users Changes to a specific user: /users/{id} |
Maximum subscription quotas: Not supported for personal Microsoft accounts like outlook.com. Not supported for Azure AD B2C tenants. A known issue for the subscription changeType. |
The following resources support rich notifications (notifications with resource data):
- Teams chat
- Teams chatMessage
- Teams channel
- Teams conversationMember
- Teams onlineMeeting *
- Teams presence *
- Teams team
Subscription lifetime
Subscriptions have a limited lifetime. Apps need to renew their subscriptions before the expiration time. Otherwise, they need to create a new subscription. For a list of maximum expiration times, see Maximum length of subscription per resource type.
Apps can also unsubscribe at any time to stop getting change notifications.
Managing subscriptions
Clients can create subscriptions, renew subscriptions, and delete subscriptions.
Creating a subscription
Creating a subscription is the first step to start receiving change notifications for a resource. The subscription process is as follows:
The client sends a subscription (POST) request for a specific resource.
Microsoft Graph verifies the request.
- If the request is valid, Microsoft Graph sends a validation token to the notification URL.
- If the request is invalid, Microsoft Graph sends an error response with code and details.
The client sends the validation token back to Microsoft Graph.
The Microsoft Graph sends a response back to the client.
The client must store the subscription ID to correlate change notifications with the subscription.
Subscription request example
POST https://graph.microsoft.com/v1.0/subscriptions
Content-Type: application/json
{
"changeType": "created,updated",
"notificationUrl": "https://webhook.azurewebsites.net/notificationClient",
"resource": "/me/mailfolders('inbox')/messages",
"expirationDateTime": "2016-03-20T11:00:00.0000000Z",
"clientState": "SecretClientState"
}
The changeType, notificationUrl, resource, and expirationDateTime properties are required. See subscription resource type for property definitions and values.
The resource property specifies the resource that will be monitored for changes. For example, you can create a subscription to a specific mail folder: me/mailFolders('inbox')/messages or on behalf of a user given by an administrator consent: users/john.doe@onmicrosoft.com/mailFolders('inbox')/messages.
Although clientState is not required, you must include it to comply with our recommended change notification handling process. Setting this property will allow you to confirm that change notifications you receive originate from the Microsoft Graph service. For this reason, the value of the property should remain secret and known only to your application and the Microsoft Graph service.
If successful, Microsoft Graph returns a 201 Created code and a subscription object in the body.
Note
Any query string parameter included in the notificationUrl property will be included in the HTTP POST request when notifications are being delivered.
Notification endpoint validation
Microsoft Graph validates the notification endpoint provided in the notificationUrl property of the subscription request before creating the subscription. The validation process occurs as follows:
Microsoft Graph encodes a validation token and includes it in a POST request to the notification URL:
Content-Type: text/plain; charset=utf-8 POST https://{notificationUrl}?validationToken={opaqueTokenCreatedByMicrosoftGraph}The client must properly URL decode the
validationTokenquery parameter provided in the preceding step, and escape any HTML/JavaScript.Escaping is a good practice because malicious actors can use the notification endpoint for cross-site scripting type of attacks.
In general, treat the validation token value as opaque, as the token format can generally change without notice. Microsoft Graph never sends any value containing HTML or JavaScript code.
The client must provide a response with the following characteristics within 10 seconds of step 1:
- A status code of
HTTP 200 OK. - A content type of
text/plain. - A body that includes the URL decoded validation token. Simply reflect back the same string that was sent in the
validationTokenquery parameter.
The client should discard the validation token after providing it in the response.
Important
If the client returns an encoded validation token, the validation fails.
- A status code of
Additionally, you can use the Microsoft Graph Postman collection to confirm that your endpoint properly implements the validation request. The Subscription Validation request in the Misc folder provides unit tests that validate the response provided by your endpoint.

Renewing a subscription
The client can renew a subscription with a specific expiration date of up to three days from the time of request. The expirationDateTime property is required.
Subscription renewal example
PATCH https://graph.microsoft.com/v1.0/subscriptions/{id}
Content-Type: application/json
{
"expirationDateTime": "2016-03-22T11:00:00.0000000Z"
}
If successful, Microsoft Graph returns a 200 OK code and a subscription object in the body. The subscription object includes the new expirationDateTime value.
Deleting a subscription
The client can stop receiving change notifications by deleting the subscription using its ID.
DELETE https://graph.microsoft.com/v1.0/subscriptions/{id}
If successful, Microsoft Graph returns a 204 No Content code.
Change notifications
With a client subscribing to changes to a resource, Microsoft Graph sends a POST request to the notification URL whenever the resource changes. It sends notifications only for changes of the type that's specified in the subscription, for example, created.
Note
If a client has multiple subscriptions that monitor the same resource and use the same notification URL, Microsoft Graph can send multiple change notifications that correspond to different subscriptions, each showing the corresponding subscription ID. There is no guarantee that all change notifications in the POST request belong to a single subscription.
Change notification example
This section shows an example of a notification for a message creation. When the user receives an email, Microsoft Graph sends a change notification as shown in the following example.
Note that the notification is in a collection represented in the value field. See changeNotificationCollection for details of the notification payload.
When many changes occur, Microsoft Graph may send multiple notifications that correspond to different subscriptions in the same POST request.
{
"value": [
{
"id": "lsgTZMr9KwAAA",
"subscriptionId":"{subscription_guid}",
"subscriptionExpirationDateTime":"2016-03-19T22:11:09.952Z",
"clientState":"secretClientValue",
"changeType":"created",
"resource":"users/{user_guid}@{tenant_guid}/messages/{long_id_string}",
"tenantId": "84bd8158-6d4d-4958-8b9f-9d6445542f95",
"resourceData":
{
"@odata.type":"#Microsoft.Graph.Message",
"@odata.id":"Users/{user_guid}@{tenant_guid}/Messages/{long_id_string}",
"@odata.etag":"W/\"CQAAABYAAADkrWGo7bouTKlsgTZMr9KwAAAUWRHf\"",
"id":"{long_id_string}"
}
}
]
}
Processing the change notification
Your process should process every change notification it receives. The following are the minimum tasks that your app must perform to process a change notification:
Your process should process every change notification it receives and send a 2xx class code. If Microsoft Graph doesn't receive a 2xx class code within 3 seconds, it will try to publish the change notification a number of times, for a period of about 4 hours; after that, the change notification will be dropped and won't be delivered. If your process consistently does not respond within 3 seconds, your notification might be subject to throttling.
If your processing is expected to take more than 3 seconds, you should persist the notification, return a
202 - Acceptedstatus code in your response to Microsoft Graph, then process the notifications. If the notification is not persisted, return a 5xx class code to indicate an error so the notification will be retried.If your processing is expected to take less than 3 seconds, you should process the notifications and return a
200 - OKstatus code in your response to Microsoft Graph. If the notification is not processes correctly, return a 5xx class code to indicate an error so the notification will be retried.Validate the
clientStateproperty. It must match the value originally submitted with the subscription creation request.Note
If this isn't true, you should not consider this a valid change notification. It is possible that the change notification has not originated from Microsoft Graph and may have been sent by a rogue actor. You should also investigate where the change notification comes from and take appropriate action.
Update your application based on your business logic.
Repeat for other change notifications in the request.
Throttling
Send a 202 - Accepted status code as soon as you receive the change notification, even before validating its authenticity. You are simply acknowledging the receipt of the change notification and preventing unnecessary retries. For most subscriptions, notifications are not subjected to any additional delay during publishing and all notifications are delivered within the SLA unless the service is experiencing an incident. However, if a subscription notification URL is slow or fails to response, notifications might be throttled for the host associated with the subscription. The following process is used to determine when to throttle and how to handle throttled endpoints.
Notifications are published using an HTTP client with a 3 second timeout. Upon completion of publishing of a notification, regardless of outcome, the total time spent attempting to publish, including network latency, is tracked for the host associated with the notification URL. If the publishing time is greater than 2900 ms, the response is considered slow. The responses are accumulated for the host and the percentage of slow responses is calculated after 100 notifications have been received. When the percentage of slow responses reaches 10%, the host associated with the notification URL is flagged as a slow endpoint. Because slow endpoints are associated with the host in the notification URL, all notifications for all subscriptions associated with the host are considered for evaluation and are subjected to throttling.
The evaluation continues in real time. If the publishing time for a host drops below 290 0ms, this non-slow response is included in the total count of responses, the percentage of slow responses is re-calculated, and the endpoint is re-evaluated. Additionally, the accumulation of responses is flushed every 10 minutes and the evaluation starts again, waiting for 100 notifications before evaluating an endpoint. Therefore, a temporary spike in network latency or publishing delay will recover after the delay is mitigated. A more persistent network latency or publishing delay greater than 2900 ms will be continually re-evaluated every 10 minutes.
While an endpoint is throttled, notifications are subjected to the following additional delays:
- Notifications are automatically offloaded to a set of workers dedicated to failed and throttled notifications and an additional delay of 10 minutes is incurred.
- Notifications are dropped if the throttled endpoint slow % is >= 15%.
- Notifications that failed to deliver due to an unsuccessful HTTP call are retried again in 10 minutes.
Code samples
The following code samples are available on GitHub.
- Microsoft Graph Training Module - Using Change Notifications and Track Changes with Microsoft Graph
- Microsoft Graph Webhooks Sample for Node.js
- Microsoft Graph Webhooks Sample for ASP.NET Core
- Microsoft Graph Webhooks Sample for Java Spring
Firewall configuration
You can optionally configure the firewall that protects your notification URL to allow inbound connections only from Microsoft Graph. This allows you to reduce further exposure to invalid change notifications that are sent to your notification URL. These invalid change notifications can be trying to trigger the custom logic that you implemented. For a complete list of IP addresses used by Microsoft Graph to deliver change notifications, see additional endpoints for Microsoft 365.
Note
The listed IP addresses that are used to deliver change notifications can be updated at any time without notice.
Latency
The following table lists the latency to expect between an event happening in the service and the delivery of the change notification.
| Resource | Average latency | Maximum latency |
|---|---|---|
| alert | Less than 3 minutes | 5 minutes |
| callRecord | Less than 15 minutes | 60 minutes |
| channel | Less than 10 seconds | 60 minutes |
| chat | Less than 10 seconds | 60 minutes |
| chatMessage | Less than 10 seconds | 1 minute |
| contact | Unknown | Unknown |
| conversation | Unknown | Unknown |
| conversationMember | Less than 10 seconds | 60 minutes |
| driveItem | Less than 1 minute | 5 minutes |
| event | Unknown | Unknown |
| group | Less than 2 minutes | 15 minutes |
| list | Less than 1 minute | 5 minutes |
| message | Unknown | Unknown |
| onlineMeeting | Less than 10 seconds | 1 minute |
| presence | Less than 10 seconds | 1 minute |
| printer | Less than 1 minute | 5 minutes |
| printTaskDefinition | Less than 1 minute | 5 minutes |
| team | Less than 10 seconds | 60 minutes |
| [todoTask][] | Less than 2 minutes | 15 minutes |
| user | Less than 2 minutes | 15 minutes |
Note
The latency provided for the alert resource is only applicable after the alert itself has been created. It does not include the time it takes for a rule to create an alert from the data.
See also
Feedback
Submit and view feedback for