Requirements

Everything you need before running the installer. The installer checks most of this for you and refuses to deploy until the blocking items pass, so nothing fails half-way.

In Databricks

An account admin login. Needed to sign in at the account level and list the workspaces in your account, so you can choose where the appliance lives.

At least one Unity Catalog enabled workspace attached to your account’s metastore. Two or more if you want to see the account-wide, multi-workspace view working.

System tables enabled. The system.* schemas must be available in your account. Four are required:

Schema	What it supplies
`system.billing`	Spend and DBU data. This is the cost floor. Without it there are no dollars to report.
`system.compute`	Cluster inventory, node types, utilization.
`system.lakeflow`	Jobs, tasks, runs, pipelines.
`system.query`	SQL statement history and warehouse attribution.

One more is optional but recommended:

Schema	What it supplies
`system.access`	Workspace names. Without it the dashboard shows workspace ids instead. Everything else still works.

Missing schemas do not fail the install. The detections that need them are skipped and named, and the rest run. See Verification checks.

A SQL warehouse in the hub workspace, for the scan to read through. If you do not have one, the installer can create a small serverless warehouse named lakemine-wh for you.

Workspace admin rights to create a schema in one catalog on the hub workspace. This is one time and is revoked by uninstalling.

A metastore admin available, if you are not one yourself. The SELECT on system.* grant is a metastore-level permission that account admin and workspace admin do not confer. The installer applies it when the person running the install has the authority, and prints the exact GRANT statements to hand over when they do not.

Nothing to prepare for the run identity. The installer creates, or reuses, a dedicated lakemine-scan service principal and grants it what the scan needs automatically.

On the machine running the installer

The Databricks CLI. This is the packaged installer’s only hard local prerequisite. The installer checks for it and shows the one-line install command for your operating system if it is missing.


# macOS
brew tap databricks/tap && brew install databricks
 
# Windows
winget install Databricks.DatabricksCLI
 
# Linux
curl -fsSL https://raw.githubusercontent.com/databricks/setup-cli/main/install.sh | sudo sh

A browser, for the OAuth sign-in steps. The installer opens your default browser so you authenticate through your normal SSO and MFA.

Outbound network access to your Databricks account and to the Lakemine portal.

In your Lakemine account

A Lakemine account at lakemine.ai , or created from the installer’s sign-in screen, which registers you through the same portal and starts a 14-day trial.

An active subscription. The installer refuses to deploy if the subscription is not active.

There is no account id or license key to type anywhere. Your identity comes from the session you sign in with, which is also what the subscription check and the metering credential are bound to, so the three cannot drift apart.

Supported clouds

Databricks on AWS, Azure, and GCP. You pick your cloud on the first step of the install, which determines the account console host the installer signs you in against.

Things that are not required

Worth stating explicitly, because these come up in review:

No agent, and no long-running Lakemine process in your environment. The only thing that runs is one scheduled Databricks job on your own compute.
No inbound network access to your workspace from Lakemine.
No standing credential held by Lakemine. Access is granted at install, in your own environment, and revoked by uninstalling.
No permissions for dashboard viewers. Group membership is the entire access control. Viewers need nothing on the catalog, the result tables, or system.*.

Quickstart takes you from download to first findings.