What UnoRouter does
A routing layer for language model APIs: developers use a single key and the OpenAI request format to call models from many labs, and failed calls are retried on another provider automatically.
UnoRouter sits between your application and the companies that serve large language models. You create one API key, point any OpenAI SDK or client at a single base URL, and choose a model by name. The same request format then works for GPT, Claude, Gemini, DeepSeek, Qwen, Mistral and others, so changing provider is a one string edit instead of a new integration, and there is a single prepaid balance instead of a stack of separate bills.
Behind the endpoint, a routing layer tracks which upstream providers are healthy for each model. When one times out, rate limits or returns errors, the request is retried on another provider that serves the same model. For an app or an agent loop this means an upstream incident usually shows up as a slightly slower answer rather than a failed call. Users who prefer a specific upstream can pin it, and the model catalog shows which providers serve each model and at what price.
Because it speaks the OpenAI protocol, UnoRouter works with tools that already accept a custom base URL: Claude Code, Cline, SillyTavern, RisuAI and plain scripts in Python or JavaScript. There is no extra SDK to install. The website also has a chat client that stores conversations in the browser rather than on the server, so chat history is not kept by the service.
The frontend, the gateway and the provider sync tooling are public on GitHub, which makes the routing and privacy behaviour something you can read rather than take on trust. A small set of models can be used for free with rate limits, which is enough to try the service or run light personal projects. Other models are billed per request from a prepaid balance, starting with a small top up, and there is no monthly subscription or seat fee.
Compared with closed routers, the main differences are the open code, the local first chat, and the focus on keeping requests alive through provider failures.
Best for
Developers and hobbyists who call several AI model providers and want one key, one balance and automatic failover.
Not for
Teams that must keep all traffic inside their own cloud account with a single provider contract.
Questions, answered by the maker
What does it do?
It gives you one API key and one OpenAI compatible endpoint for models from many providers, and retries requests on another provider when one fails.
Who is it for?
Developers, agent builders and hobbyists who use more than one model provider or want failover without writing it themselves.
How much does it cost, and is there a free plan?
Some models are free with rate limits. Other models are pay as you go from a prepaid balance, with no subscription.
How do I get started?
Sign up, create an API key, set the base URL of your OpenAI SDK or client to the UnoRouter endpoint and pick a model name from the catalog.
Which AI models does it use, and can I choose them?
Models from OpenAI, Anthropic, Google, DeepSeek, Qwen, Mistral and others. You choose the model per request by name.