Platform overview
Apps, AppKit, Lakebase Postgres, and Agent Bricks are layers of a full-stack Databricks application running inside your workspace. New to building on Databricks? See Start here first.
| Layer | What it is |
|---|---|
| Apps | The hosting layer. Your app runs as a managed workspace resource with a fixed URL, built-in auth, and managed compute. Deploy with databricks apps deploy. |
| AppKit | The TypeScript SDK for building apps on Databricks Apps. Built-in Databricks OAuth handling, pre-built React UI components (data tables, charts, dialogs), and a plugin system for connecting to workspace services. |
| Lakebase Postgres | The database layer. Managed Postgres for OLTP, co-located with your Lakehouse. Autoscales on demand, scales to zero when idle, and supports branching for development environments. |
| Agent Bricks | The AI layer. Call Knowledge Assistants, Supervisor Agents, and governed LLM endpoints via the Model Serving plugin. Query Unity Catalog tables in natural language via the Genie plugin. |
| Data Lakehouse | The analytical data tier. Unity Catalog tables, materialized views, and streaming tables populated by Lakeflow. Read with the Analytics plugin. Trigger Lakeflow Jobs from your app. |
How a request flows
A user opens the app, and Databricks Apps authenticates them via workspace SSO. The relevant AppKit plugin then queries Lakebase, a Genie space, or a serving endpoint on their behalf, using either the app's service principal or the user's forwarded token.
Where to next
These docs go deeper on each platform layer:
- Web apps (Databricks Apps): Scaffold and deploy an AppKit app (TypeScript) or a Python app on Databricks Apps.
- Lakebase Postgres: Provision a Lakebase Postgres project and connect it to an app.
- Data Lakehouse: Read Unity Catalog tables with the Analytics plugin, trigger Lakeflow Jobs, and surface data freshness.
- Agent Bricks: Governed model endpoints (AI Gateway), natural-language data queries (Genie), and custom agent endpoints (Custom agents).
- Set up your environment: Databricks CLI, agent skills, and MCP server.
- AppKit Reference: Component library, plugin API, and TypeScript SDK reference.