looplj/axonhub

AxonHub is an open-source AI gateway written in Go. It sits between your application and AI model providers, translating requests so that one SDK or API format can reach many different providers without requiring code changes. The core idea is that you point your existing OpenAI or Anthropic SDK at AxonHub instead of the real provider, and AxonHub routes the request to whichever model you have configured. Switching from one model to another means changing a setting, not rewriting your integration. The gateway supports over 100 language models across more than 10 providers. It includes built-in failover, which means that if one provider endpoint fails, AxonHub automatically routes to a working one, targeting a switch time under 100 milliseconds. Load balancing across multiple provider channels is also built in. On the observability side, every request is traced end-to-end, including timing information. The cost tracking feature records how many input, output, and cache tokens each request used, and the associated cost, so you can monitor spending per request in real time. The project also includes role-based access control for teams, with per-user quotas and data isolation. AxonHub is maintained by a single person. The README notes that the project does not cover converting subscriptions to API access, and that users should evaluate the risks of relying on a solo-maintained tool. Installation is via Docker or by compiling from source with Go. A live demo instance is publicly available with a sample account. Documentation is hosted separately and covers the API reference, tracing setup, permissions, load balancing, and cost tracking.