Try in the browser, and deploy the exact same platform in your cloud or on premises.
Data never leaves AustraliaWorking endpoint in seconds
If your data need to stay inside your trust boundary, you have three options:
Stand up Kubernetes, model serving, LLM gateway, governance and get an authority to operate. Team disappears for six months.
Buy a platform that needs thousands of professional services hours to install and run.
A browser demo today, effortless install in your cloud or hardware tomorrow. Ready in days, not months.
Run agents and apps. Secure and effortless infrastructure for the agentic era in your trust boundary.
Kubox serves open source and custom models through an OpenAI and Anthropic compatible endpoint, with named keys, spending limits and a full usage record built in. Existing code works by changing two lines.
Managed endpoints (SaaS), bring your own kubernetes or install it on bare metal, three options same user experience.
Any model, your boundary, fully governed.
Sign up, add credits, and get a working AI endpoint in about 60 seconds, served from sovereign Australian capacity through SCX. It speaks the API format your developers already use, so existing code works by changing one line.
The front door. Prove the experience before you commit.
Have a fine-tuned or custom model? A built-in guide inspects it, works out exactly what hardware it needs, estimates the monthly cost, and produces a ready-to-deploy plan.
Tell us about your modelNamed keys per team, spending limits, model-by-model access control, and a full usage record for chargeback and audit. The layer platform leads are accountable for, working identically on SCX.ai or in your own cluster.
Accountability you can hand to an auditor.
If we are not the right fit, we will let you know in the first meeting.
Try in the browser today and a simple install in your cloud or hardware tomorrow.
You don't need a PhD to understand our pricing.
Comfortable from the boardroom to terminal and every software engineering role in between.
If it runs on Kubernetes, it most likely will run in Kubox.
Use Dask, Spark and federated query engines like Trino for distributed data processing.
Data pipelines written in Python with Dagster, tracked in version control and packaged as containers.
ML models deployed and served with Ray Serve, a framework agnostic python library that supports model composition, multiplexing, and fractional GPUs.
Use Qdrant as a self-hosted open-source vector database for next-generation AI-native semantic search and meaningful information extraction from unstructured data.
Build self-hosted Retrieval Augmented Generation (RAG) pipelines to access your organisation's past and current knowledge in form of multi-modal data.
In partnership with SCX
SCX.ai runs sovereign Australian AI inference on SambaNova Reconfigurable Dataflow Unit (RDU) racks at Equinix Sydney, and powers the Kubox demo.
Book a walkthrough and we deploy your model plan on a real Kubox cluster on the call. The software arrives with a human attached.
“Regulated teams shouldn't have to choose between using AI and keeping their data inside their walls. Kubox gives them the easy path the hosted clouds offer, on infrastructure they control, with governance an auditor can trust.”
Chinkit Patel
Founder, Kubox