razorBridge — compute

razorBridge is KlusAI’s compute product. You rent GPU capacity as blades, work on them over SSH, and pay from a prepaid credit balance in euros. The same balance also pays for inference gates — an OpenAI-compatible model endpoint you can point an application at.

razorBridge is the developer-facing product in the family: most of what you do here happens through the rb command-line client or the REST API. It is built on kOS, the platform shared with Lex and Med.

Two ways to get a GPU

• Self-serve sessions — you start a blade for yourself, pick a duration, and it stops automatically when the time is up. Billed to your own credits.
• Events — an organizer schedules a workshop, lab or competition; blades are provisioned before the start time and torn down at the end. Participants join with a code and never see a bill; the organizer’s account pays.

Start here

Page	What it is for
Quickstart	Install the CLI, sign in, get onto a GPU
GPU blades and sessions	How a blade is provisioned, extended and stopped
Credits and billing	How usage draws down your balance
Inference gates	Model endpoints for applications
Run an event	For organizers: workshops, labs, competitions
Manage an event budget	For organizers: credits, estimates, guardrails
CLI reference	Every rb command, flag and config file
API reference	The REST API behind the CLI
Troubleshooting	Errors and what they mean

Live product surfaces

Current GPU availability, published hourly rates and the cost calculator are on razorbridge.eu — rates are read from the live pricing catalogue, so that page (or GET /v1/gpu-types/) is always the authority. This documentation deliberately does not repeat euro rates.

Note

Every new account receives €10 of credit when it is first created, which is enough to try a small GPU and a few inference calls. Self-serve credit purchase is not available yet — see Credits and billing for how to top up in the meantime.