CDF services have limits that protect system performance and enforce contractual terms. Understanding limits helps you plan your data architecture, troubleshoot errors, and avoid service interruptions.
Why limits matter
Limits ensure fair resource allocation across all CDF projects and maintain system stability. They also align with your contract terms, defining what resources are available to your project. When you understand your project’s limits, you can:
- Plan capacity: Estimate whether your current and planned data volumes fit within your project’s limits
- Troubleshoot errors: Identify when API errors or service failures are caused by exhausted limits
- Prevent incidents: Monitor usage and request limit increases proactively before hitting limits
Types of limits
CDF uses several types of limits to manage resources and API usage. Understanding these categories helps you identify which type of limit might be affecting your operations.
- Resource count limits: Restrict the total number of objects you can create for a given resource type. For example, CDF projects with the default configuration have a limit of 5 million data model instances.
- Consumption limits: Restrict usage by tracking how many times you call an API, how many executions you run, or how many Cognite units have been consumed. These limits are typically set for a specific time period, and the count resets at the beginning of each new period.
- Concurrency limits: Control the number of API operations that can run simultaneously on your CDF project. These limits are set per API and vary across different API endpoints.
- Rate limits: Constrain the number of API calls you can make to an endpoint over a time period. Cognite APIs have rate limits that apply both to projects and to individual service principals or user identities.
Reaching limits
When you reach or exceed a limit, CDF responds in different ways depending on the limit type:
-
Resource count limits: Operations that would create new resources fail with an error message indicating the limit has been reached. You cannot create additional resources until you delete existing ones or the limit is increased.
-
Rate and concurrency limits: For requests that exceed rate or concurrency limits, CDF returns an HTTP
429 Too many requests response status code.
For best practices on avoiding throttling due to rate or concurrency limits, see
Request throttling.
-
Consumption limits: CDF provides feedback on usage when limits are exceeded. To change consumption limits, contact your Cognite Account Team to discuss contract terms.
CDF service limits
Different CDF services have different limit types. The tables below show where to find detailed limit information for each service.
| Service | Limit types |
|---|
| Data modeling | Resource counts, property values, concurrency |
| Cognite Functions | Functions per project, CPU, RAM, timeouts |
| Data workflows | Workflows, executions, triggers |
| Atlas AI agents | Agents and tools per project |
| IAM resources | Groups, security categories |
| Service API | Limit types |
|---|
| Assets | Rate and concurrency limits |
| Events | Rate and concurrency limits |
| Files | Rate and concurrency limits |
| Records | Rate and concurrency limits |
| Streams | Rate and concurrency limits |
| RAW | Request and concurrency limits |
| Transformations | Concurrency limits |
| Time series | Aggregation limits |
| Sequences | Aggregation limits |
| Documents | Search limits |
Need higher limits?
If you frequently exceed a limit, contact Cognite Support or your Cognite Account Team to check if limits can be increased. Some limits are adjustable, while others are fixed product limits or tied to your contract terms. Last modified on February 17, 2026
