Build production AI through one API

Use ModelRail aliases to send chat, reasoning, code, and tool-calling workloads through an OpenAI-compatible chat endpoint.

How it works

• ModelRail exposes an OpenAI-compatible chat/completions interface at https://api.modelrail.dev/v1.
• Applications set model to a ModelRail alias such as modelrail-auto or modelrail-chat.
• Applications do not select raw upstream providers or provider model IDs.
• ModelRail may route or fail over requests behind the gateway. The response model is always a resolved ModelRail alias.

Compatibility

Point OpenAI-compatible clients at https://api.modelrail.dev/v1. Coverage is chat completions and the models list — not the full OpenAI API. Unknown request fields are rejected.

Make your first request

Create an API key in the dashboard, set the base URL, and send a completion with modelrail-auto.

Make your first request →