Skip to main content
Untitled oneOf schemas appear here as “Option 1”, “Option 2”, etc.For Documents, /documents/search, /documents/passages/search, and /documents/aggregate may show raw $ref text in descriptions.Use the Cognite API reference at https://api-docs.cognite.com/20230101/ for clearer labels and full text until we have fixed the upstream OpenAPI.

​
OpenAPI downloads

You can download the OpenAPI specification as JSON to quickly get started with API exploration and testing:

​
Request URLs

CDF exposes a RESTful API. Requests use a URL of this form: 
https://api.cognitedata.com/api/{version}/projects/{project}/{resource}
  • {version} — Path API version (stable REST APIs use v1). See API versions.
  • {project} — Project identifier for the resources you are accessing.
  • {resource} — Resource type (for example assets, files, events). Add query parameters or path segments as documented for each endpoint.
For HTTP methods (GET, POST, and so on), standard operation patterns (create, list, filter, retrieve, and so on), projects, and resource types, see Use the API.
Cluster base URLs and current multi-tenant IP addresses for allowlisting outbound traffic are listed under IP addresses for the API in Use the API.

​
API versions

All stable Cognite REST APIs use the v1 path prefix. Within v1, behavior is pinned with calendar versions: dates in the form YYYYMMDD passed in the Cdf-Version request header. To request a specific calendar version, set: 
Cdf-Version: <YYYYMMDD>
For example, Cdf-Version: 20230205 selects API behavior as released on that date or earlier. If you omit Cdf-Version, requests default to calendar version 20230101, so existing clients keep working as new versions are released. Breaking changes are shipped as new calendar versions; non-breaking changes apply across the API. For deprecation policy (including support for older versions after they are marked deprecated), removed legacy API versions, and detailed rules for backwards-compatible vs breaking changes, see API versions.

​
Beta versions

We do not recommend beta versions for production environments.
Beta versions preview upcoming stable features. Request them with the Cdf-Version header using a date and the -beta suffix (for example, Cdf-Version: 20230205-beta). See API versions for SDK support and lifecycle.

​
Alpha versions

Alpha versions are experimental and may change or be removed at any time. They are indicated by an -alpha suffix (for example, Cdf-Version: 20230101-alpha). Use with caution.

​
Using the Cognite API with Postman

1

Download the OpenAPI specification

Use the JSON download links in OpenAPI downloads above.
2

Import into Postman

Open Postman and select Import. In the Import dialog box, drag and drop the downloaded file or folder, or use the file picker.
3

Configure import settings

In View Import Settings:
  • Set Folder organization to Tags
  • Turn off Enable optional parameters
  • Turn on Always inherit authentication Select Import to complete the process.
4

Set up authentication

Set the Authorization type to Oauth2.0. By default, settings are for Open Industrial Data. Visit Cognite Hub to obtain credentials for use in Postman.
Open Industrial Data is a public Cognite Data Fusion project with sample industrial data. You can use it to explore the API without your own CDF project or identity provider administrator access. For an overview, see What is Open Industrial Data? on Cognite Hub.
For more information, see Getting Started with Postman.

​
Pagination

Most resource types support pagination, indicated by the nextCursor field in the response.
To retrieve additional pages:
1

Check for nextCursor

If the response includes a nextCursor, more results are available.
2

Request the next page

Pass the value of nextCursor as the cursor parameter in your next request. All other parameters must remain the same.
For more detail on paging, see Paging.

​
Parallel retrieval

Parallel retrieval improves performance for complex queries by splitting data requests into multiple partitions.
To use parallel retrieval:
1

Choose partition count

Decide how many partitions (n) to split the data set into (maximum 10 recommended).
2

Make parallel requests

For each partition m (from 1 to n), make a request to the endpoint (for example, /events) with the parameter partition=m/n.
3

Paginate within each partition

Follow the pagination process for each partition, passing the partition parameter to all subqueries.
Processing parallel retrieval requests is subject to concurrency quota limits. If you exceed the limit, you will receive a 429 (Too Many Requests) response. Use at most 10 partitions to prevent issues. Some resources enforce this limit automatically.
For more details, see the partition attribute documentation for the specific CDF resource.

​
Request throttling

If you exceed your project’s request capacity, CDF returns an HTTP 429 (Too Many Requests) status code. For rate and concurrency limits per API, throttling causes, and best practices for retry strategies, see API rate limits.
Last modified on April 23, 2026