Open source is how we build.
We don't think inference infrastructure should be a black box. The tools we build to run it, from the gateway to our kernel and optimization work, we put in the open so you can read them, run them, and make them better. No gated cores, no bait-and-switch.
Everything we maintain in the open.
Tensormux Gateway
PythonThe OpenAI-compatible inference gateway: routing, health-based failover, and observability in one self-hosted binary.
kernel-skills
TypeScriptA skill library that helps AI coding agents write, optimize, and debug high-performance CUDA and Triton kernels.
TensorPath
PythonInference-optimization control plane that picks the best GPU, backend, and quantization for a model, with a Triton kernel-gen layer.
Our flagship, the open front door to the platform.
Route, health-check, and observe your inference backends with a single config file. Run it standalone, or grow into the managed control plane when you need scale.
$ docker pull ghcr.io/krxgu/tensormux:latestWant to build with us?
Star a repo, open an issue, or send a pull request. Everything is permissively licensed, self-hostable, and free of vendor lock-in.