> 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/access-and-security.md).

# Access and Security

How MariaDB Cloud AI Retrieval authenticates requests, connects to your database, and handles the region of your data.

This page explains who can use AI Retrieval, how your applications authenticate, how AI Retrieval connects to your MariaDB service, and where your data is processed. Document knowledge bases and database knowledge bases handle the database connection differently, so each has its own section.

## Roles and Permissions

Your MariaDB Cloud role decides what you can do with knowledge bases:

| Role          | Access                                                  |
| ------------- | ------------------------------------------------------- |
| Viewer        | Can view knowledge bases. Cannot create or change them. |
| Member        | Can create, change, and delete knowledge bases.         |
| Administrator | Can create, change, and delete knowledge bases.         |
| Billing       | No access to knowledge bases.                           |

## API Keys and Authentication

Applications call a knowledge base through a REST API and authenticate with an API key. Send the key as a bearer token in the `Authorization` header, or in the `X-API-Key` header. An API key belongs to your user account, not to a single knowledge base.

To create, revoke, or delete a key, see [Connect Your Application](https://mariadb.com/docs/mariadb-cloud/cloud-ai/ai-retrieval/document-knowledge-bases.md#connect-your-application).

## Document Knowledge Bases

A document knowledge base stores its chunks and embeddings in a MariaDB service that you select or let MariaDB Cloud provision. AI Retrieval does not read any other data in that service.

### Objects That AI Retrieval Creates

For each document knowledge base, AI Retrieval creates these objects in the vector store service:

| Object        | Name                                                                             | Purpose                                                                               |
| ------------- | -------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------- |
| Schema        | `ai_retrieval_cloud_` followed by the last 7 characters of the knowledge base ID | Holds the chunks and embeddings of the knowledge base.                                |
| Database user | `mdb_ai_rag_` followed by the same 7 characters                                  | The account that AI Retrieval uses to connect. It has privileges on this schema only. |

AI Retrieval creates these objects with a service user that has permission to create users. You do not supply credentials for a document knowledge base, and the portal does not offer an option to choose the user name.

{% hint style="warning" %}
Do not delete or change these objects, and do not change the password of the database user. A knowledge base cannot ingest or retrieve documents without them.
{% endhint %}

### Network Access

AI Retrieval connects to the vector store service over a private network inside MariaDB Cloud. You do not need to add AI Retrieval to the IP allowlist of the service.

## Database Knowledge Bases

A database knowledge base reads from a MariaDB service that already holds your data. The create form pre-fills the default user of the service, and you can enter a different user name and password instead. AI Retrieval builds the semantic map of the database only from the tables and columns that this account can see.

To control what a database knowledge base can read, create a dedicated user:

1. Create a view that exposes only the data that the knowledge base should use.
2. Create a user that has privileges on the view only.
3. Enter that user when you create the knowledge base.

### Network Access for Database Knowledge Bases

A database knowledge base connects to your service from outside the network of the service. A new service accepts connections from any address (**Open to all: 0.0.0.0/0**), so you do not need to change anything. If you restrict the IP allowlist of the service with **Allow access only to the following**, add this address of AI Retrieval to it, or the connection fails:

```
34.41.226.174/32
```

To change the allowlist, open the service, select **Manage**, and then **Manage allowlist**.

## Data Residency

Document knowledge bases run in the cloud and region that you select. The vector store service must be in the same cloud and region, so the chunks and embeddings are stored there. You access a document knowledge base through the regional endpoint that the **Connect** dialog shows.

The create form marks where each embedding model runs. Text that is sent to a model is processed as follows:

| Provider                                         | Models                                                              | Processing                    |
| ------------------------------------------------ | ------------------------------------------------------------------- | ----------------------------- |
| AWS Bedrock, in the region of the knowledge base | Cohere Embed 3 English, Cohere Embed 3 Multilingual, Cohere Embed 4 | Traffic stays in the region.  |
| Google                                           | Gemini Embedding 2                                                  | Processed outside your cloud. |

## Next Steps

{% content-ref url="/pages/VvBZ0j71jH4il7tRtaBk" %}
[Document Knowledge Bases](https://mariadb.com/docs/mariadb-cloud/cloud-ai/ai-retrieval/document-knowledge-bases.md)
{% endcontent-ref %}

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

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