> For the complete documentation index, see [llms.txt](https://mariadb.com/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://mariadb.com/docs/mariadb-cloud/cloud-ai/ai-retrieval/service-quotas.md).

# Service Quotas

The default service quotas of MariaDB Cloud AI Retrieval, what happens when you reach one, and how to request a higher quota.

AI Retrieval applies service quotas to each organization. The quotas below are the defaults. To request a higher quota, contact [MariaDB support](https://mariadb.com/docs/mariadb-cloud/reference/support.md).

## Request Rates

| Quota                                                                      | Default | Scope                         |
| -------------------------------------------------------------------------- | ------- | ----------------------------- |
| Requests per second to `retrieve-chunks`                                   | 2       | Each organization in a region |
| Requests per second to `retrieve-chunks-and-generate`, including streaming | 2       | Each organization in a region |
| Requests per second to list the chunks of a knowledge base                 | 2       | Each organization in a region |

All users and API keys of an organization share these limits.

When a request exceeds a rate quota, the API returns HTTP status 429 with the error `ERR_RATE_LIMITED`. The `Retry-After` header says how many seconds to wait, and the `X-RateLimit-Limit`, `X-RateLimit-Remaining`, and `X-RateLimit-Reset` headers show the state of the limit. The error message names the quota ID to quote when you ask support for a higher quota.

## Documents and Ingestion

| Quota                                | Default | Scope                         |
| ------------------------------------ | ------- | ----------------------------- |
| Document size                        | 50 MB   | Each document                 |
| Files per upload session, in the API | 100     | Each upload session           |
| Pending upload sessions              | 5       | Each knowledge base           |
| Concurrent ingestion jobs            | 5       | Each organization in a region |
| Concurrent data source syncs         | 5       | Each organization in a region |

The portal uploads at most 5 files at a time, each up to 50 MB.

What happens at these limits:

* A document over the size limit is rejected with HTTP status 413. During a data source sync, files in the bucket that are over the size limit are skipped.
* AI Retrieval processes 5 documents of an upload session at a time. The other documents wait their turn and are not rejected.
* A data source sync over the limit is not rejected. It waits and starts by itself when a sync slot is free.

## Database Knowledge Bases

The semantic map of a database knowledge base covers up to 100 tables in each database schema and up to 1,000 columns in each table. AI Retrieval skips the rest automatically. See [The Semantic Map](https://mariadb.com/docs/mariadb-cloud/cloud-ai/ai-retrieval/database-knowledge-bases.md#the-semantic-map).

## Free Tier

During Public Preview, each organization can use up to 100 MCUs per month for AI Retrieval operations. Usage above the free tier is not available. See [Usage and Billing](https://mariadb.com/docs/mariadb-cloud/cloud-ai/ai-retrieval/usage-and-billing.md#free-tier).

## Next Steps

{% content-ref url="/pages/p23sE0E1EKazqhWykMGo" %}
[Usage and Billing](https://mariadb.com/docs/mariadb-cloud/cloud-ai/ai-retrieval/usage-and-billing.md)
{% endcontent-ref %}

<sub>*This page is: Copyright © 2026 MariaDB. All rights reserved.*</sub>

{% @marketo/form formId="4316" %}
