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

# Providers & Routing

> One gateway URL, 13 providers, your own keys. The gateway detects the right provider from the model name and forwards your request.

One base URL covers every provider Toolken supports. You keep using the SDK you already have. The gateway detects which provider to call, forwards your request, and returns the response unchanged.

## How it works

Your app sends a request to `https://gateway.toolken.ai/v1`. The gateway reads the model name, picks the right provider, attaches your provider key, and forwards the call. Your cost and usage land in the dashboard.

```
Your app  →  gateway.toolken.ai/v1  →  Provider (OpenAI, Anthropic, Gemini, ...)
```

One edge hop. No extra infrastructure. No per-provider glue code.

## Supported providers

<Table>
  <Thead>
    <Tr>
      <Th>Provider</Th>
      <Th>Name value (for `X-Toolken-Provider`)</Th>
      <Th>Auto-detected from model?</Th>
    </Tr>
  </Thead>

  <Tbody>
    <Tr><Td>OpenAI</Td><Td>`openai`</Td><Td>Yes (`gpt-`, `o1-`, `o3-`, `o4-`, `dall-e-`, `tts-`, `whisper-`, ...)</Td></Tr>
    <Tr><Td>Anthropic</Td><Td>`anthropic`</Td><Td>Yes (`claude-`) and any request to `/v1/messages`</Td></Tr>
    <Tr><Td>Google Gemini</Td><Td>`gemini`</Td><Td>Yes (`gemini-`)</Td></Tr>
    <Tr><Td>Groq</Td><Td>`groq`</Td><Td>Via `X-Toolken-Provider`</Td></Tr>
    <Tr><Td>Mistral</Td><Td>`mistral`</Td><Td>Yes (`mistral-`, `codestral-`, `pixtral-`, `ministral-`, `devstral-`, ...)</Td></Tr>
    <Tr><Td>DeepSeek</Td><Td>`deepseek`</Td><Td>Yes (`deepseek-`)</Td></Tr>
    <Tr><Td>Together AI</Td><Td>`together`</Td><Td>Via `X-Toolken-Provider`</Td></Tr>
    <Tr><Td>MiniMax</Td><Td>`minimax`</Td><Td>Yes (`minimax-`, `abab-`, ...)</Td></Tr>
    <Tr><Td>xAI</Td><Td>`xai`</Td><Td>Yes (`grok-`)</Td></Tr>
    <Tr><Td>OpenRouter</Td><Td>`openrouter`</Td><Td>No - always use `X-Toolken-Provider: openrouter`</Td></Tr>
    <Tr><Td>Cerebras</Td><Td>`cerebras`</Td><Td>Via `X-Toolken-Provider`</Td></Tr>
    <Tr><Td>Fireworks</Td><Td>`fireworks`</Td><Td>Via `X-Toolken-Provider`</Td></Tr>
    <Tr><Td>Perplexity</Td><Td>`perplexity`</Td><Td>Yes (`sonar-`)</Td></Tr>
  </Tbody>
</Table>

<Note>OpenRouter model names use `vendor/model` notation that overlaps with real provider prefixes, so auto-detection is intentionally disabled. Always set `X-Toolken-Provider: openrouter` when routing through OpenRouter.</Note>

## BYOK: you own your keys

Toolken does not pool or hold provider credentials. You send your own provider key on every request. The gateway forwards it verbatim and never stores it. You keep full control of your provider account and billing relationship.

See [Authentication](/get-started/authentication) for the two-key model.

## OpenRouter: hundreds of models through one provider

Setting `X-Toolken-Provider: openrouter` routes your request to OpenRouter, which aggregates hundreds of upstream models under a single OpenAI-compatible surface. It is a convenient way to reach models from providers not listed above or to compare outputs across providers.

## Next

<CardGroup cols={2}>
  <Card title="Choose a Provider" icon="sliders" href="/features/providers-routing/choose-provider">How the gateway picks a provider, and how to override it with `X-Toolken-Provider`.</Card>
  <Card title="Anthropic & Gemini" icon="sparkles" href="/features/providers-routing/anthropic-gemini">Using Anthropic's native Messages API and Gemini through the OpenAI-compatible endpoint.</Card>
</CardGroup>