This document gives an overview of the characteristics and common features of the API.

The commercetools platform HTTP API provides an interface for programmatic access to the data stored on the platform and associated functionality. While the API can be used directly, for experimentation or custom integration solutions, we recommend taking a look at our available client libraries and SDKs, which provide a more comfortable development experience.


In order to provide the best possible Service Level Agreement (SLA) and minimal latency, the commercetools platform API is provided in two different regions: Europe and United States.
These regions are completely isolated from each other and no data is transferred between them.

Platform accounts

Platform accounts created for one region are not valid for the other regions. A signup is required for each region individually:

  • Europe:
  • United States:

In case you need advice in which region your project should be located, please contact commercetools support.


The commercetools HTTP API is hosted at the following URLs:

  • Europe:
  • United States:

All requests to endpoints should be prepended with these URLs.

The Machine Learning endpoints use different hosts, which are listed with each endpoint.

Disallowed requests

Requests to the HTTP API are blocked if any of the following conditions are true:

  • The combination of request URL and headers is longer than about 15KB.
  • The request method does not allow a body, but the request has one.
  • The request contains an upgrade header.
  • The HTTP version is unknown.

It is strongly advised to use the host and domain names in your applications instead of hardcoding any IP address of the API hosts. IP addresses can change any time without prior notice.


The HTTP API requires a valid authorization token for all requests. For more information, see Authorization.

Some endpoints provide alternative scopes for each API operation. In addition to these, the manage_project:{projectKey} scope is always implicitly valid. This scope grants full API access for all operations within a project.


The majority of the HTTP API endpoints consume and produce standard application/json payloads. To ensure that payloads are encoded correctly, use Content-type: application/json in your HTTP header when sending POST requests.

When consuming a JSON repsonse, ignore any unrecognized fields encountered in JSON objects of responses, or deal with them in a way that does not cause the client application to crash. The API assumes clients behave properly and considers the addition of new fields to be backwards-compatible with respect to existing clients. In addition, client applications should not rely on the presence of undocumented fields. Fields returned by an endpoint that are documented can be renamed or removed at any time without prior notice.


Some resources contain an id field which uniquely identifies a resource. The ID is generated by commercetools when creating a resource.

Some resources have a string field key which also uniquely identifies a resource. The key identifier is always assigned to a resource by the client.

Update Guarantees

The API provides different update guarantees for different kinds of requests.

Strong Consistency

Each resource provides read-after-write consistency. Read-after-write consistency means that when you update an entity, wait for the successful response, and then read the same entity, you see the changes you just made.

Eventual Consistency

Some updates are not visible immediately after the update. For example, if you update a discount, updating the discount itself provides read-after-write consistency, but the prices on all affected products are updated after a delay. When you read a price immediately after the discount, you may read an the old value and not the updated one. If you wait a bit, and retrieve the price again, you will read the new value.

Similarly, after updating a product the Product Projection Search index is updated after a delay. If you retrieve a search result immediately after an update, the result may not reflect the updated product. When trying later, the search index will be updated.

These delayed updates provide an Eventual Consistency guarantee. It means that after a delay, what you can read reflects all past updates.

These delays can vary depending on the amount of data to update. We do our best to keep these delays as small as possible.

Optimistic Concurrency Control

Many API resources use optimistic concurrency control to prevent lost updates when making changes to existing data concurrently. These resources have a version attribute. When sending (partial) updates to these resources, the expected version of a resource is sent as a part of the request.

After making an update, the HTTP response body contains the new version of the resource. An API client must wait for the new version before sending follow-up requests.

Background processes and other events can also change a resource’s version number without any requests sent from an API client directly to the resource. Do not rely on a consistently incremented version number.

The API does not use ETag and If-Match HTTP headers for the purpose of optimistic concurrency control.

If a version mismatch occurs, the API returns a 409 HTTP status (Conflict) error.


All API access is over HTTPS. API authorization uses ↗ OAuth2 bearer tokens. See the section on authorization for details.

All HTTPS traffic is secured through Transport Layer Security (TLS). The relevant certificates are signed by trusted authorities. At any point in time, we will rely on a trusted root certificate. It is the API user’s obligation to keep their systems updated regarding SSL/TLS capabilities as well as trusted root certificates. The server certificate as well as intermediary certificates of the certificate chain are transmitted as part of the TLS handshake.

Please note that it is actually possible to connect to the API via HTTP (without SSL/TLS). You will, however, receive a 404 error including an HTML body response on every attempt.

It is not necessary to actually support the full set of SSL/TLS features listed here (or stop supporting other features), especially the ciphersuites, as long as one valid combination exists. The TLS handshake includes a “negotiation phase” that chooses the best combination that both client and server support.

Be aware that our certificates may at times support more features in transition periods. However, only the specifications listed below are reliable. You must not rely on other SSL/TLS features even if they currently work.

  • Versions: TLSv1.2

Note: With this setup, we are following recommendations for modern compatibility by the Mozilla Foundation. Our API is, however, only bound to what is detailed out above.

Query Features

Common query features include sorting, paging, and reference expansion. Certain aspects of these features are standardized across all parts of the API as described in the following sections.

In order to filter results, Query Predicates allow to express complex queries on many resources.

Single Resource Query Response

Query responses for single resources only return a single, top-level JSON object.


For query responses of requests supporting paging via limit and offset, the following common top-level fields are returned:

  • offset - Number
    The offset supplied by the client or the server default. It is the number of elements skipped, not a page number.
  • limit - Number
  • count - Number
    The actual number of results returned in results.
  • total - Number - Optional
    The total number of results matching the query.
    This field is returned by default.
    For improved performance, calculating this field can be deactivated by using the query parameter withTotal=false.
  • results - Array of results
  • facets - Object containing FacetResults - Optional
  • meta - Object containing supplementary information about the results - Optional


A query endpoint that supports sorting does so through the sort query parameter. The provided value must be a valid sort expression.
The default sort direction is ASC. The allowed sort paths are typically listed on the specific query endpoints.

Here are some examples of sort expressions:

  • name desc
  • dog.age desc

If multiple sort expressions are specified via multiple sort parameters, they are combined into a composed sort where the results are first sorted by the first expression, followed by equal values being sorted according to the second expression, and so on.

Note: The sorting is case sensitive.


Queries on collections of resources can optionally provide paging functionality through limit and offset query parameters, thereby allowing clients to request a certain page of the complete query result.


The maximum number of results to return from a query can be set using the limit query parameter. Allowed is a value between 1 and 500. The default limit on most query endpoints is 20.


The offset into the results matching the query can be set using the offset query parameter. An offset of 0 is the default value indicating that no results should be skipped.

Note: When iterating on all elements from an endpoint (for example to import/export all products), using offset for that is not appropriate:

  • when an element is inserted at the same time, request with next offset value might return already processed elements.
  • using a large offset can lead to performance problems.

In cases where you want to iterate over all elements while avoiding offset you can follow this alternative approach instead:

  • Query for a limited result in which the elements are sorted by their ID.
  • Iterate with the same query, but starting with an ID that is greater than the last ID from the previous step in order to get the next chunk of elements.

The following pseudo-code demonstrates how to iterate on all elements returned by an endpoint, using chunks from 100 elements at each iteration:

lastId = null
continue = true
while (continue) {
if (lastId == null)
response = client.get(endpoint + "?withTotal=false&limit=100&sort=id+asc")
response = client.get(endpoint + "?withTotal=false&limit=100&sort=id+asc&where=id%3E" + lastId) // "id%3E" is "id >" encoded
results = response.results
continue = (results.size == 100)
lastId =
// do something with result

Resource Timestamps

Many resources exposed on the API contain the createdAt and the lastModifiedAt timestamps. Those timestamps are automatically set by the server on creation or on modification of the resource. They can be used as predicates and for sorting in queries.

Reference Expansion

Reference expansion is a feature of the resources listed in the table below that enables clients to request server-side expansion of Reference resources, thereby reducing the number of required client-server roundtrips to obtain the data that a client needs for a specific use-case.

Reference expansion can be used when creating, updating, querying, and deleting these resources.

Endpoints supporting Reference Expansion on creation, update, query and delete
Carts CartDiscounts Categories Channels
Customers CustomerGroups DiscountCodes State
Inventory Messages Orders Products
ProductDiscounts ProductProjections ProductTypes Reviews
ShippingMethods TaxCategories Types Zone
API Extensions

The JSON snippet below is taken from an Example-Product representation that contains two references: one to an Example-Product-Type and one to an Example-TaxCategory:

"id": "<example-product-id>",
"version": 4,
"productType": {
"typeId": "product-type",
"id": "<example-product-type-id>"
"taxCategory": {
"typeId": "tax-category",
"id": "<example-tax-category-id>"

Without Reference Expansion you’ll get such a representation as response from the query endpoints.

With Reference Expansion you’ll get the expanded objects embedded into the product by an additional obj field as shown in the following JSON snippet:

"id": "<example-product-id>",
"version": 4,
"productType": {
"typeId": "product-type",
"id": "<example-product-type-id>"
"obj": {
"id": "<example-product-type-id>",
"version": 4,
"name": "Example Product Type",
"description": "example Product Type to showcase Reference Expansion",
"classifier": "Complex",
"attributes": [
"taxCategory": {
"typeId": "tax-category",
"id": "<tax-category-id>"
"obj": {
"id": "<tax-category-id>",
"version": 2,
"name": "Standard tax category",
"rates": [
"name": "5% US",
"amount": 0.05,
"includedInPrice": false,
"country": "US",
"id": "zJJ5KIGH"
"name": "19% MwSt",
"amount": 0.19,
"includedInPrice": true,
"country": "DE",
"id": "893f_nXJ"

All query endpoints that support reference expansion provide an expand query parameter which can be used to specify one or more expansion paths.

Expansion paths define the full path to the reference in a resource. Expansion path format:

  • refFieldName - expands a reference field named refFieldName located at the root of the resource
  • refArrayFieldName[*] - expands all references contained in the array of references named refArrayFieldName
  • refArrayFieldName[0] - expands the first reference in the array of references named refArrayFieldName
  • objFieldName.refFieldName - expands the reference refFieldName contained in the object objFieldName
  • objArrayFieldName[*].refFieldName - expands the reference refFieldName in all objects contained in the array objArrayFieldName

The query for the Example Product introduced before - including both expansion paths - would look like this:


A more complex example is to expand all categories within a found product. The expansion path would be: masterData.current.categories[*].

The query would thus look like this:


The expansion path can also point to references in expanded objects.

Note that expansion of references does not cause requests to fail if not all of the references can be expanded (e.g. because the referenced resource cannot be found). The server will try to expand as much as possible but never beyond what the client requested. It is up to the client to decide which referenced resources it deems critical for completing a specific use-case, e.g. by either treating it as an error or by simply not rendering / returning the missing information.

Data Erasure of Personal Data

In some countries, particularly in the EU, you are required by a law such as the GDPR to erase personal data of a customer on request, also known as “Right to be Forgotten”.

A default DELETE request may not clean up all data, both visible at the HTTP API layer (e.g. personal data may be part of messages) and invisible (the commercetools platform internally keeps logs for some time, to reconstruct data in case of faulty system behavior).

Endpoints that store personal data therefore offer a parameter for DELETE requests called dataErasure. If set to true, the commercetools platform guarantees that all personal data related to the particular object, including invisible data, is erased, in compliance with the GDPR. You are, however, responsible for identifying and deleting all objects that belong to a customer, and deleting them.

The parameter is available for:

Personal data MUST NOT be stored in objects other than the ones listed above.

Partial Updates

For partial updates to existing resources, a common approach used by the HTTP API is to accept a custom patch format in the form of a list of resource-specific actions that need to be processed. The request method for partial update requests is usually POST, but PATCH might also be supported in the future.

The general request representation (patch format) of a partial update request is as follows:

  • id - String - Optional if part of the URL The ID of the targeted resource.
  • version - Number - Required if the targeted resource is versioned The expected current version of the resource on which to apply the changes.
  • actions - Array of resource-specific actions - Required

Example Request

"id": "a136fd9e-bc29-4d05-813f-350b248aefd8",
"version": 1,
"actions": [
"action": "setTitle",
"title": "My title"
"action": "setText",
"text": "My text"

The concrete actions that can be used on a specific resource are documented together with the resource.

All actions provided in a specific resource’s patch format are generally freely combinable in a single request. Furthermore, such update requests only succeed or fail as a whole, i.e. unless explicitly stated otherwise, the result of all actions sent in the request is applied to a resource or none at all, if any of the actions fail.

Not all updates or state transitions of a resource may be freely composable with one another in a single HTTP request. For those updates, other endpoints may be provided to handle only a specific update or state transition in a single HTTP request.

Correlation ID

A Correlation ID is a unique identifier of a single action or event (such as checking out a cart), allowing to trace the action across multiple systems. For example, a mobile app makes a request to its backend, that backend makes another request to the commercetools platform, and the commercetools platform makes yet another request to an API Extension. If the mobile app sets a Correlation ID, and the backend forwards it to the commercetools platform, it’ll also be forwarded to the API Extension. Therefore, the action can be followed through the logs of all four systems.

Correlation IDs are strings that may only contain alphanumeric characters, underscores and hyphens and have a length of 8 to 256 characters. UUIDs can be used to ensure uniqueness.

The HTTP header X-Correlation-ID is used to propagate the Correlation ID on HTTP requests and responses. If a client does not provide a Correlation ID for an HTTP request, the commercetools platform generates one.

When contacting our support regarding a specific request, please also include the Correlation ID as it allows us to look up the request more quickly.