Olyx started as an internal tool inside an AI-powered construction cost platform I was building. I was routing between a frontier model for CAD extrapolation and a private model for routine tasks. It worked — but I had no way to know if a routing change was actually cheaper, or whether it had quietly started sending PII to the wrong endpoint. Cost tracking was bolted on after the fact. Auditing production meant digging through provider logs and application logs separately and piecing it together manually. At some point I was maintaining infrastructure instead of building the product. I scrapped the library-based approach — Ruby, Python, TypeScript, three codebases for the same logic — and went CLI-first. The CLI speaks to the control plane directly. Your application points at one endpoint. The language your stack is written in stops mattering. One binary. Every stack supported from day one. Policy before the call. Every decision recorded from the first request. No library to install. Early access is open — link in the post. Happy to answer anything on the routing engine, CLI design, or what you're running into in your own stack.

Olyx

One integration. Every AI request governed and traced.

One integration. Every AI request governed and traced.