Skip to main content
Uniblock Gateway is an OpenAI-compatible LLM router. You change one base URL and one API key, keep the OpenAI SDK you already use, and call models across providers on a single prepaid balance. We hold the provider accounts and meter the requests, so you never hold a provider credential.

​
The two lines that change

Point your existing client at https://ai.uniblock.dev and give it a Uniblock key. Nothing else in your code moves.
Python
Model ids are provider/model, so the id you send says which provider serves it. A bare model name with no provider is refused rather than routed for you.

​
What it costs

Credit is prepaid. A platform fee of 2.5% is added when you buy credit, and requests take the provider’s rate at face value with no markup on top. Buying 100 credits charges 102.50 and grants 100 credits, all 100 of which are spendable. The 2.5% rate is the current rate rather than a permanent one, and the rate that applies to a purchase is fixed when you start checkout.

​
Start here

Quickstart

Get a key, set the base URL, and send your first chat completion.

OpenAI compatibility

Which endpoints exist today, what is additive, and what is refused.

Authentication

Key format, the Bearer header, shown-once keys, and revocation.

API reference

The wire contract: endpoints, headers, and the error envelope.

​
What is not here

There is no self-hosted or downloadable version of the gateway. The product is hosted: we hold the provider accounts, meter the requests, and sell prepaid credit, and none of that works from a copy running somewhere else. There is no static list of providers or models on this site, because that set is a live fact rather than a documentation fact. Ask GET /v1/models or read the model list in the dashboard instead of a page that would go stale.