Databricks AI Dev Kit

MCP (Model Context Protocol) server for building Databricks projects with AI coding assistants like Claude Code and Cursor.

Overview

High-level, AI-assistant-friendly tools for Databricks operations. Organized by product line for scalability.

Architecture

databricks-mcp-core/              # Pure Python library
├── unity_catalog/                # Catalogs, schemas, tables
├── compute/                      # Execution contexts
├── spark_declarative_pipelines/  # Pipeline management & workspace files
├── agent_bricks/                 # Agent Bricks (future)
└── dabs/                        # DAB generation (future)

databricks-mcp-server/            # MCP protocol wrapper
├── server.py                    # FastAPI + SSE
└── tools/                       # MCP tool wrappers

Installation

# Using uv (recommended)
uv venv
uv pip install -e databricks-mcp-core
uv pip install -e databricks-mcp-server

# Or using pip
pip install -e databricks-mcp-core
pip install -e databricks-mcp-server

# Configure authentication (use your Databricks profile)
export DATABRICKS_CONFIG_PROFILE=your-profile
# or set DATABRICKS_HOST and DATABRICKS_TOKEN

Quickstart

Using the MCP Server with Claude Code

1. Install packages (see above)
2. Configure authentication (uses DEFAULT profile from ~/.databrickscfg if not set)
3. Configure Claude Code MCP settings (see below) - no need to manually start server!
4. Ask Claude to list your catalogs, create tables, run pipelines, etc.

Using the Core Library Directly

from databricks_mcp_core.unity_catalog import catalogs, schemas, tables
from databricks.sdk.service.catalog import ColumnInfo, TableType

# List catalogs (returns List[CatalogInfo])
all_catalogs = catalogs.list_catalogs()
for catalog in all_catalogs:
    print(catalog.name)

# Create a table (returns TableInfo)
table = tables.create_table(
    catalog_name="main",
    schema_name="default",
    table_name="my_table",
    columns=[
        ColumnInfo(name="id", type_name="INT"),
        ColumnInfo(name="name", type_name="STRING")
    ],
    table_type=TableType.MANAGED
)

All functions use the official databricks-sdk and handle authentication automatically.

Available Tools (33)

Unity Catalog (11):

• Catalogs: list_catalogs, get_catalog
• Schemas: list_schemas, get_schema, create_schema, update_schema, delete_schema
• Tables: list_tables, get_table, create_table, delete_table

Compute (4):

• create_context, execute_command_with_context, destroy_context, databricks_command

Spark Declarative Pipelines (15):

• Pipeline Management (9): create_pipeline, get_pipeline, update_pipeline_config, delete_pipeline, start_pipeline_update, validate_pipeline, get_pipeline_update_status, stop_pipeline, get_pipeline_events
• Workspace Files (6): list_pipeline_files, get_pipeline_file_status, read_pipeline_file, write_pipeline_file, create_pipeline_directory, delete_pipeline_path

Synthetic Data Generation (3):

• get_synth_data_template, write_synth_data_script_to_workspace, generate_and_upload_synth_data

Usage with Claude Code

Add to .claude/.mcp.json:

{
  "mcpServers": {
    "ai-dev-kit": {
      "command": "uv",
      "args": ["run", "python", "-m", "databricks_mcp_server.stdio_server"],
      "transport": "stdio"
    }
  }
}

Note: The server runs automatically when Claude needs it - no manual startup required!

If not using uv, use:

{
  "mcpServers": {
    "ai-dev-kit": {
      "command": "python",
      "args": ["-m", "databricks_mcp_server.stdio_server"],
      "transport": "stdio"
    }
  }
}

Then ask Claude to interact with your Databricks workspace!

Documentation

• databricks-mcp-core README - Core package details
• databricks-mcp-server README - Server configuration

License

© 2025 Databricks, Inc. All rights reserved. The source in this notebook is provided subject to the Databricks License [https://databricks.com/db-license-source].

📄 Third-Party Package Licenses

© 2025 Databricks, Inc. All rights reserved. The source in this project is provided subject to the Databricks License [https://databricks.com/db-license-source]. All included or referenced third party libraries are subject to the licenses set forth below.

Package	License	Copyright
databricks-sdk	Apache License 2.0	Copyright (c) Databricks, Inc.
pydantic	MIT License	Copyright (c) 2017 Samuel Colvin
fastapi	MIT License	Copyright (c) 2018 Sebastián Ramírez
uvicorn	BSD 3-Clause License	Copyright © 2017-present, Encode OSS Ltd.
sse-starlette	BSD 3-Clause License	Copyright (c) 2020, sysid
python-dotenv	BSD 3-Clause License	Copyright (c) 2014, Saurabh Kumar