AI-Assisted development

Prerequisites

• Node.js environment
• Databricks CLI: install and configure it according to the official tutorial.
• A new Databricks app with AppKit installed. See Bootstrap a new Databricks app for more details.

AppKit integrates with AI coding assistants through the Model Context Protocol (MCP).

Installing MCP server

To install the Databricks MCP server for your preferred AI assistant, run:

databricks experimental apps-mcp install

MCP capabilities

The MCP server exposes the following capabilities for AI assistants:

• Data exploration: Query catalogs, schemas, tables, and execute SQL
• CLI command execution: Deploy and manage apps, and run other workspace operations
• Workspace resource discovery: Inspect and navigate workspace resources

Example prompts

Here are some basic examples you can use to explore your app and workspace:

1. Creating a new basic app

Create a new Databricks app that displays a dashboard of the users table in main.default.

1. Introspecting a table schema

Show me the schema of the users table in main.default.

1. Creating a new query

Create a new query to find users created in the last 7 days.

1. Deploying an app

Deploy the app to Databricks with the name "my-app".

Possibilities are virtually endless. Ask your AI assistant to help you with your app and workspace.

LLM resources

AppKit provides specialized guidance files to help AI coding assistants work more effectively with the projects.

If you're building applications using AppKit packages, reference the llms.txt file.