A filled environment template
Every variable the install reads, with the values the CLI renders for itself already marked, so the only ones left to chase are the ones your administrators send back.
Self-host
Run it inside your own tenant
Kubernetes in your own tenant, or local Docker on one machine. The starter pack is the document your operators would otherwise have to assemble themselves.
Studio runs as containers on one machine under local Docker, or on an existing AKS cluster under Kubernetes on Azure. Either way it sits inside your own perimeter, against DuckDB (local), Microsoft Fabric, and MotherDuck, using your keys and your model endpoints.
The install itself is short: one installer, one command. What takes the time is everything that has to be true before it — app registrations, capacity, workspace roles, a GitHub App, outbound network access. Those sit with people who are not you, so the first week is mostly requests in flight. The starter pack turns each of them into a single page you can forward, addressed to the person who holds the right it needs.
What is in the starter pack
A sizing worksheet
What the cluster has to carry. The core profile needs 8 vCPU across at least two nodes and full monitoring needs at least 10 — measured on a real install, where two four-vCPU nodes sat at 88–95% CPU allocation and could not schedule one of Studio's own components.
A prerequisites checklist
Your deployment style and your data platform together decide which asks apply, and no combination needs all of them. The checklist says which can go out on day one and which are waiting on a value from another.
The Microsoft Fabric admin requests
Six asks, each tagged with the right it needs — tenant, capacity, or workspace administration — so you route each one to whoever holds it, rather than sending the whole page to a single administrator and waiting.
Get the starter pack
The pack is free. Your platform capacity, your cluster and your model calls run on your own accounts, and what a licence costs is a conversation. Local Docker is the smaller setup, not a cloud-free one: every combination needs an Azure AI Foundry resource. The side-by-side comparison sets out what each deployment style needs before day one.
Not wired The form is markup only in this prototype. Nothing is sent, and the starter pack is not delivered from here yet. Until it is, hello@acceleratedata.ai reaches a founder directly.