Gateway

OpenAI-compatible endpoints for supported frontier models. Direct OpenAI, Anthropic, and Google routing with streaming, tool calling, and chat fallbacks.

What is the opengateway Gateway?#

The opengateway Gateway is a single OpenAI-compatible HTTP endpoint that routes requests to OpenAI, Anthropic, or Google based on the model ID you pass. One API key and one Chat Completions request shape work across supported providers.

What it is#

The Gateway is a single HTTP endpoint that speaks OpenAI's API and forwards your request to the correct model at the correct provider. It supports Chat Completions, streaming, tool calling, provider-aware structured output parameters, and request-level chat fallbacks.

What it solves#

Many applications want to use Claude Sonnet for reasoning, GPT-4o-mini for summarization, and Gemini for long-context jobs. Without a gateway, that usually means three SDKs, three billing accounts, three retry strategies, and a scramble every time one of the providers has a bad hour.

With the opengateway Gateway, it is three model IDs on the same endpoint.

One endpoint, everything#

POST https://api.opengateway.ai/v1/chat/completions
POST https://api.opengateway.ai/v1/responses          # OpenAI models only

Model families today#

opengateway connects directly to the three companies shipping the frontier models. There is no reseller in the middle, and no hosting-vendor markup. When you call openai/gpt-4o, you are hitting the model OpenAI built.

Owner	Model family	Notes
OpenAI	GPT series (gpt-4o, gpt-4o-mini, o-series)	Chat, Responses, tools, streaming.
Anthropic	Claude series (Sonnet, Opus, Haiku)	Chat Completions shape with tool translation.
Google	Gemini series (2.5 Pro, Flash)	Chat Completions shape with multimodal inputs.

The full catalog is on the Models page.

Key features#

Chat Completions

Drop-in OpenAI-compatible API.

Streaming

SSE token streams, normalized across providers.

Fallbacks

Request-level fallback models for Chat Completions.

Tool Calling

OpenAI tool format, translated for supported providers.

Pricing model#

The Gateway charges provider pass-through cost plus a small platform fee (a few percent). Billing records provider cost, platform fee, and total cost.

See Billing for the details.

Where to use something else#

Model training and fine-tunes happen at the provider. Once the fine-tune exists, you can route to it through opengateway like any other model.
Prompt versioning is handled by your own prompt store.
End-user chat UIs are your job. opengateway is the platform you build on.