Bring your own model

By default, Mintlify selects the models that power the assistant and agent. To use a specific model with your own provider account, supply an API key and choose the model. Mintlify then runs inference against your provider. One configuration covers every AI product on a project: the assistant, agent, and automations. Bring your own model applies to a single project, not to your whole organization. If your organization has more than one project, configure each one separately. Projects you don’t configure use Mintlify’s default models.

​

Supported providers

• OpenRouter
• Anthropic
• OpenAI
• Amazon Bedrock

With any supported provider, the assistant, agent, and automations all use your key and your model. Each provider supports a fixed set of models. The model menus list only the models available for the provider you select. For Amazon Bedrock, the available models depend on the region you select.

​

Configure your model

Editors and admins can configure the model.

1. Select the project you want to configure.
2. Navigate to the API keys page in your dashboard.
3. In the AI model section, enable Bring your own model.
4. Select a provider. For Amazon Bedrock, also select a Region.
5. Select an Assistant model.
6. Optional: Select an Agent model to run the agent and automations on a different model. If you leave it empty, they use the assistant model.
7. Enter your provider API key. For Amazon Bedrock, use an Amazon Bedrock API key. Mintlify doesn’t support AWS access keys or IAM role credentials.
8. Click Save.

Mintlify stores your key encrypted and never displays it again after you save it.

​

How credits work

AI usage on a project with bring your own model enabled does not consume Mintlify credits. Your provider bills you directly for inference, and those charges do not appear in your Mintlify usage. Set any spending limits with your provider. Credits pool at the organization level. Other projects in the same organization still draw from that shared balance at the standard rate for their AI usage.

​

End-user identifiers

Mintlify sends a hashed identifier for the end user with each request to OpenAI, Anthropic, and OpenRouter. If your provider’s safety systems flag abusive traffic, the provider can block that one user instead of your whole API key. The identifier is an HMAC hash of your project and the user’s IP address, conversation, or Mintlify account. It does not contain the raw IP address, email, or other personal details. The same user gets the same identifier across requests, so you can match flagged activity in your provider’s dashboard.

​

Rotate or remove your key

Mintlify reuses your saved key across models for the same provider on that project. Enter a key again only when you rotate it, switch providers, or configure another project. To rotate your key, enter the new key in the API key field, then click Save. To stop using your own model, disable Bring your own model. Mintlify deletes the stored configuration and key for that project, and its AI products return to Mintlify’s default models. AI usage on the project then consumes credits at the standard rate. Other projects keep their own configurations.