> ## Knowledge Base Index
> Fetch the complete knowledge base index at: https://helpdesk.gbnetwork.com/sitemap.xml
> Use this file to discover available pages before exploring further.
> Pure-Markdown content can be obtained by appending a '.md' suffix to the content URLs listed in the sitemap (without the trailing slash).

# How to Add a Custom AI Provider in OpenClaw

This guide walks you through adding a **custom AI provider** that is not listed in the standard `setup-ai` command using the `openclaw configure` interactive tool.

This method is useful for providers like:

* NVIDIA NIM
* Together AI
* Deepseek
* Fireworks AI
* Any OpenAI-compatible endpoint

In this example, we will add **NVIDIA NIM** with the model `nvidia/nemotron-3-super-120b-a12b`.

### Prerequisites

* **Active OpenClaw VPS -** Your VPS is running and accessible.
* **SSH Access -** You can connect to your VPS via SSH.
* **API Key -** You have an API key from your custom provider (if required).
* **Provider Endpoint -** You know the API endpoint URL.
* **Model Name -** You know the exact model identifier.

### Step-by-Step Guide

1. SSH into Your VPS.

2. Run the Configure Command:

```
openclaw configure
```

3. When asked `"Where will the Gateway run?"`, select **Local** (since this will be implemented on your VPS).
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-1_17qioe9.png =500xauto)

4. When asked `"Select sections to configure"`, select **Model** (because we want to add an AI provider/model).
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-2_k0zhxs.png =500x251)

5. When asked `"Model/auth provider"`, you may choose any model provider you want. For this guideline, you may select **Custom Provider** (since the provider is not listed in the default options).
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-3_17mxjda.png =500x704)

6. When asked `"API Base URL"`, enter your provider's API endpoint URL.

* For NVIDIA NIM:
```
https://integrate.api.nvidia.com/v1
```

![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-4_n5c36e.png =500xauto)

7. When asked `"How do you want to provide this API key?"`, select **Paste API key now**.
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-5_16mpkre.png =500xauto)

8. When asked `"API Key (leave blank if not required)"`, paste your actual API key.

* For NVIDIA, it typically starts with `nvapi-`

![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-6_1gimzfs.png =500x88)

9. When asked `"Endpoint compatibility"`, select **OpenAI-compatible** (most custom providers use this format).
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-7_fgx3f3.png =500x116)

10. When prompted for `"Model ID"`, enter the exact model identifier provided by your AI provider.
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-8_1uravoj.png =500x91)

⚠️ Since most providers offer multiple models, make sure you are using a valid one. If you encounter a `verification failed` error, you can run the following command to check which models are available and compatible:

```
curl <API Base URL>/models -H "Authorization: Bearer <API Key>" -s | jq '.data[].id'
```

* `<API Base URL>` with your AI provider’s API endpoint
* `<API Key>` with your actual API key from the provider dashboard

Example:
```
curl https://integrate.api.nvidia.com/v1/models -H "Authorization: Bearer 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" -s | jq '.data[].id'
```

11. When asked `"Endpoint ID"`, you can accept the auto-generated value or modify it.
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-9_1uekub6.png =500x78)

12. When asked `"Model alias (optional)"`, you can enter a shorter, friendlier name for this model.
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/oc-10_1hg640o.png =500x100)
13. Once completed, the system will display a success message confirming that your custom provider has been successfully configured. The full configuration result is shown below:
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/another-full_4lbxf3.png =500x634)

14. In your chat interface (Telegram, WhatsApp, or web chat), type `/new` to start a fresh session with your new custom model.
![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/chatbox-nvidia2_1muepno.png =900x427)

| You have complete add custom AI Provider.

---

### Verification

Check the configuration file:

```
cat /opt/openclaw/data/.openclaw/openclaw.json | grep -A 20 "custom"
```

![](https://storage.crisp.chat/users/helpdesk/website/-/6/7/9/3/67938060e89a4400/verification-custom_11lvftt.png =500x674)

### Need Further Assistance?

If you face any issues or need assistance, don’t hesitate to reach out — **our support team is always ready to help!**

🔧 **Need help?** [**Submit a Support Ticket**](https://secure.gbnetwork.com/submitticket.php)  
💬 **Chat with us on Live Chat via our** [**website**](https://www.gbnetwork.my/)