> ## Documentation Index
> Fetch the complete documentation index at: https://docs.e6data.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Models

> Enable model providers in the Model Registry, and choose models per conversation and per workflow step.

export const DocMeta = ({type, applies, owner}) => {
  const LABELS = {
    tutorial: "Tutorial",
    "how-to": "How-to guide",
    explanation: "Explanation",
    reference: "Reference"
  };
  return <div className="e6-meta">
      <span className={type === "how-to" || type === "tutorial" ? "e6-eyebrow" : "e6-eyebrow e6-eyebrow--quiet"}>{LABELS[type] || type}</span>
      {applies ? <span>Applies to: {applies}</span> : null}
      {owner ? <span>Owner: {owner}</span> : null}
    </div>;
};

<DocMeta type="how-to" applies="All deployments · for operators" />

AgentFlow isn't tied to one model vendor.

## Supported providers

The Model Registry supports:

* Anthropic
* OpenAI
* Amazon Bedrock
* OpenRouter
* OpenAI-compatible proxy endpoints, including self-hosted models behind a proxy or gateway

Enable at least one model before users start. Model calls come only from the default worker pods; the API gateway and sandbox pods never call a model.

## Where models are chosen

| Where | How |
| - | - |
| Chat | Users pick a model per conversation, or follow the current default. |
| Workflow steps | Pin a model on `guided`, `agent`, `subworkflow`, `map` and `action` steps. Other steps inherit the run model. |

Use per-step selection to reserve a more capable model for heavy reasoning and a cheaper one for routine extraction. Pins on step types that never call a model are rejected.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.