UnoRouter

UnoRouter started because we were paying four AI vendors, holding four keys and maintaining four SDK integrations for what was really one job: send a prompt, get an answer. So we built a single gateway in front of all of them. You get one API key and one OpenAI-compatible endpoint. Models from OpenAI, Anthropic, Google, DeepSeek, Qwen, Mistral and others sit behind it, and moving from one to another means changing the model string, nothing else. When a provider has an outage, routing shifts traffic to a healthy one automatically, so your app keeps answering. It plugs straight into existing code, Claude Code, Cline, SillyTavern and other OpenAI-compatible clients. The whole stack is open source on GitHub, and the built-in chat is local-first: your conversations stay in your browser. Pricing: a free tier with free models to start, then pay as you go. No subscription.