AlienGiraffe + Databricks
Databricks governs the data that reaches it. AlienGiraffe covers the data that can't: queried inside your network, under the policies you keep in Unity Catalog.
Where each tool fits
Copy it, federate it, or query it in place.
Three ways to reach on-prem data from Databricks. Pick the one that matches the data.
Lakeflow Connect
Copy it in.
- Runs in
- Databricks, in the cloud
- What leaves your network
- A full copy, kept in sync
- Asks of the source
- Archive-log mode, supplemental logging, and LogMiner grants on the primary
- Use it for
- Data that's allowed to live in the cloud
Lakehouse Federation
Query it from the cloud.
- Runs in
- Databricks, in the cloud
- What leaves your network
- Every result set, on every query
- Asks of the source
- A stored login and a network path from the cloud
- Use it for
- Data that may pass through the cloud
Alien Giraffe
Query it where it lives.
- Runs in
- A sandbox inside your network
- What leaves your network
- No raw rows. Only what a request approves
- Asks of the source
- A read-only login
- Use it for
- Data that can't leave, even for a moment
From Databricks' documentation, November 2026: Oracle CDC limitations, Oracle setup, Federation networking. Not sure which applies? Take the five-question fit check.
With Unity Catalog
One place to decide access, for data in both worlds.
Unity Catalog stays where access is decided. Alien Giraffe enforces it at the source and sends back only what's allowed.
Unity Catalog decides
The grants you already keep say who may ask for which data.
The sandbox loads the slice
Inside your network, only the columns and rows that policy allows. The source is read once.
Databricks gets the answer
Aggregates and masked results reach notebooks, dashboards, and agents. Raw rows stay put.
Sources, untouched
Sandbox, per request
Starts on approval, next to the source. Loads only the approved slice.
Approved consumers
AlienGiraffe Connector · your VM or Kubernetes
Databricks + Unity Catalog
Outside the line
→ in: Who may ask for what
← out: Aggregates and masked results
Raw rows don't cross the line. Not to us, not to a cloud.
Alongside Databricks today
Start with the sources Databricks can't reach.
Teams on Databricks run Alien Giraffe next to it, with approval rules kept as code.
- Request, approve, sandbox, and query for Oracle, Postgres, MySQL, SQL Server, Snowflake, S3, and log files over SSH
- Approval rules as code, by requester role, including auto-approve
- Results in the explorer, or through the API and MCP for apps and agents
Design partner program
Bring Unity Catalog policy to data that can't move.
We're building the Unity Catalog integration with a small group of design partners. You shape what Unity Catalog passes in and what comes back out.