databricks-solutions/ai-dev-kit

Databricks Toolkit for Coding Agents provided by Field Engineering

What it solves

It bridges the gap between general-purpose AI coding assistants (like Cursor, Claude Code, or GitHub Copilot) and the specific complexities of the Databricks platform. It provides the necessary context, patterns, and executable tools that these assistants need to effectively build Spark pipelines, manage Unity Catalog governance, and deploy ML models without the user having to manually explain Databricks-specific APIs and best practices every time.

How it works

The kit consists of four composable components that can be used independently or together:

  • MCP Server: Implements the Model Context Protocol to expose over 50 executable tools to AI assistants, allowing them to perform actions directly on Databricks.
  • Skills: A collection of markdown-based guides that teach AI assistants specific Databricks patterns and best practices.
  • Core Library: A Python library (databricks-tools-core) providing high-level functions for common tasks like executing SQL.
  • Builder App: A full-stack web application with a chat UI that integrates these tools for a visual development experience.

Who it’s for

Developers and data engineers who already use their own preferred AI-powered IDEs or CLI tools (such as Cursor, Windsurf, or Gemini CLI) but want to integrate deep Databricks expertise and automation into those workflows.

Highlights

  • Broad Tool Support: Compatible with a wide range of AI assistants including Claude Code, Cursor, Copilot, and Gemini CLI.
  • Extensive Toolset: Provides 75+ executable tools for building everything from AI/BI dashboards to RAG-based knowledge assistants.
  • Flexible Deployment: Can be installed as a project-level configuration, a global setup, or deployed as a standalone web app.
  • Genie Code Integration: Includes specific workflows to install skills directly into Databricks workspaces for use with Genie Code.