MCP Tools

The AI Dev Kit MCP server gives Claude Code direct, real-time access to your Databricks workspace. While skills teach Claude how to build things, MCP tools let it do things — run queries, create pipelines, deploy apps, manage governance, and more.

How It Works

The MCP (Model Context Protocol) server runs locally and proxies authenticated requests to the Databricks REST API. When Claude Code needs to interact with your workspace, it calls the appropriate MCP tool — no manual API calls or SDK scripting required.

Claude Code  →  MCP Server (local)  →  Databricks REST API  →  Your Workspace

Configuration

Add the MCP server to your Claude Code configuration:

{
  "mcpServers": {
    "ai-dev-kit": {
      "command": "uvx",
      "args": ["databricks-ai-dev-kit"],
      "env": {
        "DATABRICKS_HOST": "https://your-workspace.cloud.databricks.com",
        "DATABRICKS_TOKEN": "your-personal-access-token"
      }
    }
  }
}

See Configuration for alternative auth methods (CLI profiles, OAuth M2M, multiple workspaces).

Tool Categories

SQL & Execution

Run SQL queries, execute Python, and manage compute resources. Browse tools →

Pipelines & Jobs

Create and manage Spark Declarative Pipelines and orchestration jobs. Browse tools →

Unity Catalog & Governance

Manage catalogs, schemas, tables, grants, tags, and data sharing. Browse tools →

Dashboards & Genie

Build AI/BI dashboards, manage Genie spaces, and define metric views. Browse tools →

Vector Search

Create endpoints and indexes, query vectors, and manage embeddings. Browse tools →

Apps & Lakebase

Deploy Databricks Apps and manage Lakebase databases, branches, and syncs. Browse tools →

Serving & Models

Monitor model serving endpoints and manage AI agents. Browse tools →

Compute & Workspace

Manage clusters, volumes, file uploads, and workspace resources. Browse tools →

Skills vs. MCP Tools

	Skills	MCP Tools
What they do	Teach Claude how to approach a task	Execute actions against your workspace
When they activate	Automatically, based on your prompt	Called by Claude when it needs to interact with Databricks
Example	Spark Declarative Pipelines skill teaches pipeline design patterns	`create_or_update_pipeline` actually creates the pipeline
Analogy	A reference book	A remote control

In practice, skills and MCP tools work together. When you ask Claude to “build a streaming pipeline,” the SDP skill guides the design while create_or_update_pipeline, execute_sql, and other MCP tools handle the execution.

All Tools at a Glance

Category	Tools
SQL & Execution	`execute_sql`, `execute_sql_multi`, `execute_databricks_command`, `run_python_file_on_databricks`, `get_best_warehouse`, `list_warehouses`, `get_best_cluster`
Pipelines & Jobs	`create_or_update_pipeline`, `create_pipeline`, `get_pipeline`, `find_pipeline_by_name`, `start_update`, `get_update`, `stop_pipeline`, `update_pipeline`, `get_pipeline_events`, `delete_pipeline`, `manage_jobs`, `manage_job_runs`
Unity Catalog	`manage_uc_objects`, `get_table_details`, `manage_uc_grants`, `manage_uc_tags`, `manage_uc_monitors`, `manage_uc_sharing`, `manage_uc_storage`, `manage_uc_connections`, `manage_uc_security_policies`
Dashboards & Genie	`create_or_update_dashboard`, `get_dashboard`, `delete_dashboard`, `publish_dashboard`, `create_or_update_genie`, `get_genie`, `delete_genie`, `ask_genie`, `migrate_genie`, `manage_metric_views`
Vector Search	`create_or_update_vs_endpoint`, `get_vs_endpoint`, `delete_vs_endpoint`, `create_or_update_vs_index`, `get_vs_index`, `delete_vs_index`, `query_vs_index`, `manage_vs_data`
Apps & Lakebase	`create_or_update_app`, `get_app`, `delete_app`, `create_or_update_lakebase_database`, `get_lakebase_database`, `delete_lakebase_database`, `generate_lakebase_credential`, `create_or_update_lakebase_branch`, `delete_lakebase_branch`, `create_or_update_lakebase_sync`, `delete_lakebase_sync`
Serving & Models	`list_serving_endpoints`, `get_serving_endpoint_status`, `query_serving_endpoint`, `manage_ka`, `manage_mas`
Compute & Workspace	`list_clusters`, `get_cluster_status`, `start_cluster`, `get_best_cluster`, `get_current_user`, `manage_workspace`, `upload_file`, `upload_folder`, `upload_to_volume`, `download_from_volume`, `list_volume_files`, `get_volume_file_info`, `create_volume_directory`, `delete_volume_file`, `delete_volume_directory`, `delete_tracked_resource`, `list_tracked_resources`