SubRouter: One API for every AI model, with smart routing

Hi Product Hunt — I built SubRouter after repeatedly dealing with different SDKs, credentials, quotas, billing rules and failure modes for every AI provider. The goal is simple: let teams connect GPT, Claude, Gemini and other models through one gateway while retaining control of routing, usage, pricing and deployment. SubRouter started from the open-source One API ecosystem and evolved into a self-hosted platform with protocol conversion, weighted routing, automatic retries, cache-aware billing, dashboards and fine-grained model and token permissions. I’d especially value feedback on provider coverage, routing policies and the operational features teams need for production AI workloads. The project is open source, so you can inspect the code, self-host it and contribute.