> ## Documentation Index
> Fetch the complete documentation index at: https://docs.paywalls.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Model providers

The model provider is the underlying AI service that runs the model (for example, OpenAI, Azure OpenAI, Anthropic). The default provider and model are configured in the [Dashboard](https://dashboard.paywalls.ai), which controls available models, credentials, quotas, and routing behavior.

<Tabs>
  <Tab title="Default mode">
    **BYOK (bring your own key)** — Connect any OpenAI-compatible AI provider you already use. Simply add your API key in the Dashboard and you're ready to go.

    **Supported providers include:**

    * OpenAI
    * Anthropic
    * OpenRouter
    * Together AI
    * Any OpenAI-compatible endpoint

    **How it works:** You pay your AI provider directly for usage. Paywalls handles billing your end users and passes requests through to your provider.
  </Tab>

  <Tab title="Shared mode">
    Choose between two options:

    **Option 1: BYOK (bring your own key)**

    * Connect any OpenAI-compatible provider (OpenAI, Anthropic, OpenRouter, Together AI, or custom endpoints)
    * Add your API key in the Dashboard to get started
    * You pay your provider directly; Paywalls bills your users

    **Option 2: Built-in provider (zero setup)**

    * Use Paywalls.ai's built-in provider powered by OpenRouter
    * **No API keys, configuration, or provider accounts required**
    * Start using AI models immediately without any additional setup
    * End users' shared wallet covers the provider cost automatically
    * Paywalls splits the charge to preserve your margin
  </Tab>
</Tabs>

### Model selection & routing

* You choose a model id (e.g., `openai/gpt-4o-mini`, `openrouter/llama-3.1-70b`). We normalize requests and route to the right upstream.
* You can switch providers or models at any time without changing your integration.
* You can also set up custom routing rules to use different providers/models for different users, teams, or use cases (coming soon).