Databricks

概要

This project implements an MCP server that converts complex Databricks Unity Catalog (UC) and lineage metadata into compact Markdown summaries, specifically optimized for consumption by Large Language Models (LLMs) and agents. It provides a small REST adapter with environment-driven authentication (PAT or OAuth client-credentials) and exposes a concise set of async tools for agents to discover catalogs/schemas, inspect table metadata and lineage, and execute read-only SQL queries. The goal is to enhance automated data discovery, SQL generation, and pipeline analysis by providing AI agents with clear, structured data context directly from Unity Catalog.

主な機能

• Offers tools to fetch schemas, tables, and detailed table information
• Executes read-only Spark SQL queries against Databricks SQL Warehouse
• 0 GitHub stars
• Designed for agent composition with small, deterministic tools
• Converts UC and lineage metadata into LLM-optimized Markdown summaries
• Provides environment-driven authentication (PAT or OAuth client-credentials)

ユースケース

• Automated data discovery and exploration by AI agents
• Improved SQL generation and query precision using UC metadata
• Lineage-aware reasoning for pipeline analysis and validation
• Programmatically finding candidate tables and columns for data retrieval