LunarGate

LunarGate is a self-hosted AI gateway for teams whose apps use multiple model providers. It gives those apps one OpenAI-compatible endpoint and keeps routing, retries and fallback rules in your infrastructure. The gateway is open source under the MIT license and runs on macOS and Linux, with a container image also available.

It connects to Ollama for local models, plus OpenAI, Anthropic, DeepSeek and other OpenAI-compatible backends. Your apps can keep the same client integration when you change providers. Complexity-aware routing assigns simpler requests to cheaper models and sends demanding work to stronger ones; weighted and conditional rules give teams control over how requests reach each provider.

Retries, fallbacks and circuit breakers handle upstream failures centrally, rather than leaving each app to implement them. LunarGate supports streaming, tool calling and embeddings, along with Chat Completions and Responses APIs. Some API operations depend on the selected provider's capabilities.

Observability is optional. The gateway doesn't send ordinary prompts or responses to LunarGate's dashboard by default, though requests still go to the model provider you choose. Ollama can keep model processing local; cloud providers receive the requests routed to them. Optional observability runs in the EU and requires an account. Teams can export metrics without sharing prompt or response text, or enable that visibility selectively for debugging.