Completions API Providers

Yozakura works with any service that exposes an OpenAI completions API compatible endpoint. You just need the provider's completions API URL, a bearer/auth token if the provider requires one, and a model ID.

This document lists providers that are known (or reasonably believed) to work. Shrug emoji means "should work, but hasn't been tested".

If you test one of the shrug emoji'd providers, please let me know your results so I can update this document. You can contact via GitHub or Discord.

Similarly, if you use a provider that is not listed below, let me know and I'll add it.

Provider	Completions API URL	Tested?	Why it's good	Setup instructions/hints
OpenRouter	https://openrouter.ai/api/v1/chat/completions	✅	Access to a huge number of models with pay-as-you-go pricing.	Covered in detail in API Setup. Bearer/Auth Token is your OpenRouter API key, Model is any model ID from the models page.
Demon Route	https://api.demonroute.com/v1/chat/completions	🤷	Character-focused and uncensored fine-tunes (Dolphin, MythoMax, abliterated Llama/Qwen builds) that larger aggregators don't carry — a natural fit for NPC chat in arbitrary fictional worlds. Privacy-first: prompts are not trained on or logged, and accounts can be funded anonymously with crypto instead of an email and card. Prepaid, pay-as-you-go.	Sign up at demonroute.com and create an API key. Bearer/Auth Token is your Demon Route key, Model is any model ID from the models page. Lighter models suit the many background calls (memory processing, NPC turns); stronger ones suit main chats. See how to use different models per prompt.
KoboldCpp	http://127.0.0.1:5001/v1/chat/completions	✅	Run models on your own machine. Free, and nothing leaves your computer.	Run a model locally instead of using a cloud provider. Covered in the Local LLM section of API Setup — Yozakura's default Completions API Setup values already match KoboldCpp.