> ## Documentation Index
> Fetch the complete documentation index at: https://www.agent37.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Custom provider
> Set up OpenClaw with a custom model provider in Agent 37
Follow the steps below or watch the video walkthrough:
## Video Walkthrough
**Before you begin**:
* An active [Agent 37 account](https://www.agent37.com/dashboard)
* A running OpenClaw instance in Agent 37
* The base URL for your custom provider
* An API key if your provider requires authentication
* A model ID supported by that provider
## When to use a custom provider
Use a custom provider when:
* your preferred vendor is not in the built-in provider list
* you use an OpenAI-compatible proxy or gateway
* you route traffic through a private or self-hosted endpoint
* you want multiple external providers behind one custom base URL
## Step 1: Gather the provider details
Before you open the setup flow, make sure you have:
1. The provider base URL, for example `https://your-provider.example.com/v1`
2. The API compatibility type:
* OpenAI-compatible
* Anthropic-compatible
* Unknown, if you want OpenClaw to auto-detect
3. The API key, if required
4. The model ID you want to use
5. An endpoint ID or label so you can recognize this provider later
## Step 2: Open your Agent 37 instance
1. Go to the [Agent 37 dashboard](https://www.agent37.com/dashboard) and sign in.
2. Open an existing instance, or create a new one.
3. Open **Terminal**.
## Step 3: Start the configuration flow
In Terminal, run:
```bash theme={null}
openclaw configure
```
When prompted, choose:
1. **Where will the gateway run?** → `Local`
2. **Select Section to Configure?** → `Model`
3. **Model / Auth Provider** → `Custom Provider`
## Step 4: Enter the custom provider details
OpenClaw will guide you through the custom provider setup. You will be asked to:
1. **API Base URL** → your provider’s base API URL, such as: `https://your-provider.example.com/v1` Example: `https://api.minimax.io/v1`
According to OpenClaw’s onboarding flow, **Custom provider** is the correct choice for endpoints that are not listed directly, including providers that expose OpenAI-compatible or Anthropic-compatible APIs.
Next, complete the remaining prompts:
2. **How do you want to provide this API key now?** → `Paste API key now`
3. **Endpoint-compatibility**
* Choose `OpenAI-compatible` if your provider exposes an OpenAI-style /v1 API
* Choose `Anthropic-compatible` if it uses Anthropic’s Messages API
* Choose `Unknown` if you want OpenClaw to auto-detect compatibility
4. **Enter API key** → paste the API key you copied from your provider
5. **Model ID** → enter the exact upstream model name Example: `Minimax-M2.7`
6. **Alias (optional)** → enter a friendly display name for the model
7. **Endpoint ID** → enter a short provider identifier such as: `myproxy` or `minimax`
If your provider claims OpenAI compatibility, start with the OpenAI-compatible option first. If requests fail, verify the exact API format your provider supports.
## Step 5: Continue and verify the model
After you finish the prompt flow, run:
```bash theme={null}
openclaw models list
```
You should see the custom provider model available in the output. If you selected it as the default, it should be marked accordingly.
Your custom model should appear as:
`/`
Example:
`minimax/Minimax-M2.7`
If you want to make it your default model, run:
```text theme={null}
openclaw models set minimax/Minimax-M2.7
```
## Optional: Understand the underlying config
OpenClaw stores custom providers in the model configuration. A custom provider entry typically includes fields like:
* `baseUrl`
* `apiKey`
* `api`
* `models`
For example:
```json theme={null}
{
"models": {
"mode": "merge",
"providers": {
"custom-proxy": {
"baseUrl": "http://localhost:4000/v1",
"apiKey": "YOUR_API_KEY",
"api": "openai-completions",
"models": [
{
"id": "llama-3.1-8b",
"name": "Llama 3.1 8B"
}
]
}
}
}
}
```
“OpenAI-compatible” does not always mean full compatibility with every OpenClaw feature. Some providers only support basic chat-completions payloads and may not fully support tools, reasoning controls, or advanced response formats.
## Troubleshooting tips
* Confirm that the base URL includes the correct API path, often `/v1`
* Make sure the model ID exactly matches the provider catalog
* Recheck whether the endpoint is OpenAI-compatible or Anthropic-compatible
* Verify that your API key has active billing or access
## What to do next
Use the built-in OpenAI flow if you do not need a custom endpoint.
Use the Anthropic flow if you want the native Claude provider setup.