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

# Models & providers — Power agents with any AI model

> Power stagewise AI coding agents with any model. Use the stagewise Account, import provider subscriptions, or connect local models.

stagewise connects to large language models through **providers**. A provider hosts models — Anthropic, OpenAI, Google, DeepSeek, Z.AI, MiniMax, or any custom endpoint. Within each provider, you choose which **model** to use.

## Models vs providers vs endpoint modes

| Concept           | What it means                                                                                            |
| ----------------- | -------------------------------------------------------------------------------------------------------- |
| **Provider**      | The service that hosts models (Anthropic, OpenAI, Google, DeepSeek, etc.)                                |
| **Model**         | The specific LLM you use (Claude Sonnet, GPT-5, Gemini Flash, etc.)                                      |
| **Endpoint mode** | How requests reach the provider — through your stagewise Account, your own API key, or a custom endpoint |

## Three ways to power your agents

```mermaid theme={null}
graph TD
    A[Agent selects a model] --> B{Which provider?}
    B -->|Anthropic| C{Endpoint mode?}
    B -->|OpenAI| D{Endpoint mode?}
    B -->|Google| E{Endpoint mode?}
    C -->|stagewise Account| F[Managed inference service]
    C -->|Import subscription| G[api.anthropic.com]
    C -->|Custom endpoint| H[Your endpoint]
    D -->|stagewise Account| F
    D -->|Import subscription| I[api.openai.com]
    D -->|Custom endpoint| H
    E -->|stagewise Account| F
    E -->|Import subscription| J[Google AI API]
    E -->|Custom endpoint| H
```

Each provider is configured independently. Mix and match as needed.

### 1. stagewise Account

Start immediately with a curated library of pre-configured models. No API keys, no setup — just sign in and choose a model. Usage is billed through your stagewise account.

Best for: getting started fast, trying different models, not managing multiple API subscriptions.

### 2. Import your subscriptions (BYOK)

Already paying for Claude, GPT, Gemini, Kimi, Qwen, DeepSeek, GLM, or MiniMax? Bring your existing API keys and use them directly in stagewise. You control billing, rate limits, and data routing.

Best for: teams with existing provider contracts, maximizing subscriptions you already pay for.

[Set up BYOK →](/reference/byok-setup)

### 3. Custom & local providers

Connect any service that speaks a standard API — including models running on your machine:

* **Local inference**: Ollama, LM Studio, vLLM — zero external API calls
* **Cloud platforms**: Azure OpenAI, AWS Bedrock, Google Vertex AI
* **Enterprise proxies**: LiteLLM, any OpenAI-compatible gateway

Best for: air-gapped environments, data privacy, fixed-cost local inference.

[Connect a custom provider →](/reference/custom-providers)

## Built-in providers

stagewise supports these providers out of the box:

| Provider      | Available models           |
| ------------- | -------------------------- |
| Anthropic     | Claude Opus, Sonnet, Haiku |
| OpenAI        | GPT-5 series, GPT-5 Codex  |
| Google        | Gemini 3 series            |
| Moonshot AI   | Kimi K2 series             |
| Alibaba Cloud | Qwen 3 series              |
| DeepSeek      | DeepSeek V-series          |
| Z.AI          | GLM series                 |
| MiniMax       | MiniMax M-series           |

## Mixing modes

Each provider is configured independently. You can use your stagewise Account for OpenAI, your own Anthropic key, and a local Ollama endpoint for Google — all at the same time. The agent uses whichever mode is active for the provider of the currently selected model.

## Switching models

Switch models from the selector in the chat sidebar. Each agent instance can use a different model — run one agent on a fast model for simple tasks, another on a powerful model for complex architecture work.

## Custom models

Add models not in the default list — fine-tuned models or models served by your custom endpoint. [Add custom models →](/reference/custom-models)

## What's next

<CardGroup cols={2}>
  <Card icon="key" href="/reference/byok-setup" title="Import your subscriptions">
    Step-by-step guide to bringing your own API keys.
  </Card>

  <Card icon="cpu" href="/reference/custom-providers" title="Custom providers">
    Connect local models via Ollama or any custom endpoint.
  </Card>

  <Card icon="sliders-horizontal" href="/reference/models-and-providers" title="Models & providers reference">
    Full configuration reference for every provider and setting.
  </Card>
</CardGroup>