Skip to content
OneAPIDocs
ModelsOpen console

Start here

The OneAPI API

One key and one base URL in front of many model providers. Keep the SDK your code already uses, change the base URL, and pick a model by its ID.

https://api.oneapi.id/v1
Key header
Authorization: Bearer or x-api-key
Body
JSON, in the format of the path you call
Billing
Per token, from prepaid credit or a plan

How a request is routed

Two parts of a request matter to the gateway, and they are independent. The path decides which wire format you send and get back. The model field decides which model answers. A Claude model can be called through the OpenAI Chat Completions path, and a GPT model through the Anthropic Messages path; the gateway converts between formats on the way in and out.

The path decides the request and response format. The model field decides which model answers. Every listed model works on every path.

Conversion covers messages, system prompts, tools and streaming. A feature that one format has and the other cannot express, such as a provider-hosted tool, is refused with 400 unsupported_parameter instead of being dropped silently. See Errors.

Pick an endpoint

Call the path that matches the client you already have. There is no need to rewrite request bodies.

If your code usesCallReference
OpenAI SDK, chat.completionsPOST /v1/chat/completionsChat Completions
OpenAI SDK, responsesPOST /v1/responsesResponses
Anthropic SDK, messagesPOST /v1/messagesMessages
Anything that lists modelsGET /v1/modelsModels
Checking your balanceGET /v1/meMe

Most tools that let you set an OpenAI base URL work unchanged. Tools built on the Anthropic SDK need the base URL without /v1; Official SDKs shows both.

Where to go next