# Add a model provider

At the end of this guide your workspace can use models from a provider you pay directly, with your own API key, beside or instead of the models that came with it. You can add **OpenAI**, **Anthropic (Claude)**, **OpenRouter**, **Azure OpenAI** deployments, or any other OpenAI-compatible endpoint as a **Custom provider**. An administrator with permission to create LLM providers can do it.

> **Note:** Bringing your own provider depends on your plan; see [Pricing](/pricing). When your plan does not include it, **Add provider** is not offered, and the **Models** tab explains why.

## Before you begin

- A Volobox account with permission to create LLM providers.
- An API key from the provider. For Azure OpenAI, your resource endpoint, an API key and the names of the deployments you already created in Azure.
- The key stored in a secret scope, or permission to create a scope or key from the provider form.

## Add OpenAI, Anthropic or OpenRouter

1. Open **Admin** on the left activity rail, then **Settings**, and in **Organization Settings** open **AI**, then **Models**.
2. Select **Add provider**.
3. Choose **OpenAI**, **Anthropic (Claude)** or **OpenRouter** from the list. Volobox fills in the endpoint for you. You can change the display name.
4. Select the **Secret scope** and **Secret key** that hold your API key. To store a new key without leaving the form, use **Add new** beside either picker.
5. Select **Create**. The provider's **Models** tab opens.
6. Select **Discover models** to load the provider's models, then turn on the ones your agents should use.

> **Tip:** For OpenRouter, keep each discovered model ID unchanged, such as `anthropic/...`. Coding sessions rely on it to route the model through OpenRouter.

![The Add LLM provider dialog in Organization Settings, AI, Models: choose OpenAI, Anthropic (Claude), Azure OpenAI, OpenRouter or a custom provider, and Volobox fills in its connection settings.](https://volobox.io/docs/images/ai-provider-add.light.webp)

*The Add LLM provider dialog in Organization Settings, AI, Models: OpenAI, Anthropic (Claude), Azure OpenAI, OpenRouter and Custom provider.*

## Add an Azure OpenAI deployment

1. In **AI**, **Models**, select **Add provider** and choose **Azure OpenAI**.
2. Enter your resource endpoint and API key, or select an existing secret. The endpoint looks like one of these; Volobox adds the API path itself:

   ```text
   https://my-resource.openai.azure.com
   https://my-resource.services.ai.azure.com
   ```

3. Save, then select **Add deployment** on the **Models** tab.
4. Enter the exact deployment name from Azure, which can differ from the underlying model name. This registers a deployment you already have; it does not create one in Azure.
5. Select the API your deployment supports: **Responses API**, **Chat Completions** or **Embeddings**. Turn on image support only if the model supports it.

> **Note:** Azure uses deployment names instead of **Discover models**. This flow works with public Azure endpoints and API keys; Microsoft Entra ID and managed identity authentication are not supported here.

## Add another OpenAI-compatible endpoint

1. In **AI**, **Models**, select **Add provider** and choose **Custom provider**.
2. Enter its connection settings. Choose **No credentials required** for an endpoint that needs no key; otherwise select the secret that holds it.
3. Select **Create**, then **Discover models** on the **Models** tab.

Advanced provider options stay editable on the provider's **Connection** tab after you create it. A blank **Base URL** uses the OpenAI endpoint, not a provider-specific default.

## Check that it works

- In the provider list, open the provider's action menu and select **Test connection**. It checks that Volobox can reach the provider's model catalog.
- For an Azure deployment, select **Test deployment**. It sends a small request that the provider bills, and confirms the deployment can answer. **Test connection** alone does not prove that.
- The models you turned on appear in the model pickers in agent chats and coding sessions.

![Organization Settings, AI tab, Models: the LLM providers list with Anthropic, OpenAI and a turned-off Azure OpenAI deployment, each with its address, model catalog mode, whether it is enabled and how many of its models are enabled, and the Add provider button.](https://volobox.io/docs/images/ai-providers.light.webp)

*The LLM providers list in Organization Settings, AI, Models: each provider with its base URL, model catalog, whether it is enabled and how many of its models are on.*

## Troubleshooting

| Symptom | Fix |
| --- | --- |
| **Add provider** is not offered | Your plan does not include your own providers, or you lack permission to create LLM providers. See [Pricing](/pricing). |
| A provider is marked **Paused by plan** | It was added on an earlier plan. Its settings, models and prices are kept; it comes back as it was when the plan includes your own providers again. |
| **Discover models** is missing for Azure OpenAI | Azure lists only the deployments you add. Use **Add deployment** with the exact deployment name. |
| **Test connection** passes but an Azure model does not answer | **Test connection** checks the catalog only. Use **Test deployment**, and check the deployment name and the API you selected. |
| A model has no cost estimate in reports | Prices are not needed to use a model. To get estimates for providers that do not publish prices, such as Anthropic, enter them on the **Model prices** tab. |

## Next steps

- [Deployment and models](/docs/deployment-and-models)
- [AI agents](/docs/agents)
- [Coding sessions](/docs/coding)
- [Security and administration](/docs/security-and-administration)
