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

# Endpoints

> The request formats the AI Gateway accepts, how to name a model, and streaming.

<Info>This is in development. Details may change before it is available.</Info>

The AI Gateway speaks the request formats your clients already use. Change the base URL and the key; the request and the answer keep their usual shape.

| Format | Base URL | Main paths |
| - | - | - |
| OpenAI | `https://api.altrum.ai/v1` or `https://api.altrum.ai/openai/v1` | `chat/completions`, `responses`, `models` |
| Anthropic | `https://api.altrum.ai/anthropic` | `v1/messages` |
| Google Gemini | `https://api.altrum.ai/genai` | `v1beta/models/{model}:generateContent`, `:streamGenerateContent` |
| Cursor | `https://api.altrum.ai/cursor` | The paths Cursor calls |

Any format can reach any supported provider. An OpenAI-format request can be answered by an Anthropic model, and Altrum converts the request and the answer.

## Name a model

Name the provider and the model: `openai/gpt-4o-mini`, `anthropic/claude-sonnet-4-5`, `gemini/gemini-2.5-flash`, `bedrock/...`, `vertex/...`, `azure/...`. A model name without a provider is matched to a provider your organization has credentials for.

`GET /v1/models` lists the models your key may use.

## Streaming

Every format streams as it does with the provider directly. Guardrails on answers can delay the first words: see [Guardrails at runtime](/guardrails/at-runtime#streaming).

## Not available

The AI Gateway does not handle embeddings, images, audio, video, batch, files, containers or realtime requests yet, and a stored response cannot be fetched, listed or deleted through it. Requests to those paths are answered as an unknown path.


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