DevOps & Config

Workspace & Authentication

Skills: databricks-config MCP Tools: manage_workspace, get_current_user

Which Databricks workspace am I currently connected to?

List all available Databricks workspace profiles I can connect to.

Switch my active workspace to the staging profile.

Authenticate to a new Databricks workspace at
https://my-workspace.cloud.databricks.com

Show me the current authenticated user and their permissions.

Databricks Asset Bundles (DABs)

Skills: databricks-bundles, databricks-agent-skill-databricks

Create a new Databricks Asset Bundle project structure with dev, staging, and
prod targets.

Add a pipeline resource to my existing databricks.yml bundle configuration that
deploys to different catalogs per environment.

Add a job resource to my bundle that runs a notebook on a schedule, with
environment-specific parameters for dev vs prod.

Create a bundle configuration that deploys an AI/BI dashboard, a pipeline, and a
job as a complete analytics package.

Set up a bundle with variable substitution so the catalog name, warehouse ID,
and notification emails differ per target environment.

Configure bundle permissions so developers can manage resources in dev, but only
the service principal can deploy to prod.

Validate my bundle configuration and show me any errors or warnings before
I deploy.

Show me how to structure a bundle that uses the Direct Deployment Engine for
faster deployments.

Jobs & Workflow Orchestration

Skills: databricks-jobs, databricks-agent-skill-databricks-jobs MCP Tools: manage_jobs, manage_job_runs

Creating Jobs

Create a Databricks job that runs a notebook at
/Workspace/Users/me/etl_notebook every day at 6 AM UTC.

Create a multi-task job with this DAG:
1. Task "extract" runs a Python notebook
2. Task "transform" runs after extract completes
3. Task "load" runs after transform completes
4. Task "validate" runs after load completes
Configure it to retry failed tasks up to 2 times.

Create a job that triggers on file arrival in /Volumes/main/raw/incoming/ and
processes new files automatically.

Create a job that triggers whenever the table main.bronze.events is updated.

Set up a job with a for-each task that iterates over a list of region codes and
runs a parameterized notebook for each region.

Managing & Monitoring Jobs

List all jobs in my workspace and show their last run status.

Trigger a run of the job named "daily_etl" and monitor it until completion.

Show me the status of the latest run for job "nightly_aggregation" — did it
succeed or fail?

Check all currently running job runs and show their progress.

Cancel the currently running instance of job "backfill_historical".

Job Configuration

Update the job "daily_etl" to add email notifications on failure to
team@company.com.

Create a job that uses serverless compute with a Python wheel task that runs
my_package.main.

Set up a job with continuous scheduling that restarts automatically when
it finishes.

Python SDK & Databricks Connect

Skills: databricks-python-sdk MCP Tools: execute_databricks_command, run_python_file_on_databricks

Write a Python script using the Databricks SDK to list all clusters, find any
that have been idle for more than 2 hours, and terminate them.

Create a Python script that uses Databricks Connect to run a Spark job
remotely — read from a Delta table, transform the data, and write results back.

Write a Python script using the SDK to create a new Unity Catalog schema, create
a table with a defined schema, and insert sample data.

Build a Python automation that uses the Databricks SDK to:
1. List all jobs that failed in the last 24 hours
2. Collect their error messages
3. Generate a summary report

Write a Python script to manage model serving endpoints — list all endpoints,
check their status, and query one with test input.

Run a local Python file on my Databricks cluster that processes data with
PySpark and writes results to Unity Catalog.

Compute Management

MCP Tools: list_clusters, get_best_cluster, start_cluster, get_cluster_status, execute_databricks_command, run_python_file_on_databricks, upload_file, upload_folder

List all clusters in my workspace and show which are running, terminated,
or pending.

Start the cluster "dev-cluster" and wait for it to be ready.

Find the best available cluster and run this Python code on it:
from pyspark.sql import functions as F
df = spark.table("main.default.my_table")
df.groupBy("category").agg(F.count("*").alias("count")).show()

Upload my local project folder ./src/ to the Databricks workspace at
/Workspace/Users/me/my_project/.

Check the status of cluster "shared-analytics" — is it running and how long has
it been up?

Documentation Lookup

Skills: databricks-docs

Look up the documentation for configuring Auto Loader with schema evolution.

What are the supported file formats for the COPY INTO command?

Show me the documentation for Databricks OAuth token federation with
GitHub Actions.

What are the configuration options for serverless SQL warehouses?

Look up the latest documentation on Unity Catalog system tables — what tables
are available and what data do they contain?