Databricks CLI

Databricks CLI

Comprehensive reference for the Databricks CLI (v0.205+). Covers authentication, all command groups, flags, and common patterns.

Command syntax

databricks <command-group> [<command>] [<subcommand>] [args] [--flags]

Get help at any level with -h or --help.

Authentication

OAuth U2M (interactive, recommended)

# Workspace-level
databricks auth login --host https://<workspace>.cloud.databricks.com

# Account-level
databricks auth login --host https://accounts.cloud.databricks.com --account-id <id>

Saves credentials to ~/.databrickscfg as a named profile. Tokens auto-refresh and expire in under an hour.

OAuth M2M (service principals)

Add to ~/.databrickscfg:

[my-sp-profile]
host = https://<workspace>.cloud.databricks.com
client_id = <service-principal-client-id>
client_secret = <service-principal-oauth-secret>

Credential resolution order

1. Bundle settings files (if running from bundle directory)
2. Environment variables (DATABRICKS_HOST, DATABRICKS_TOKEN, etc.)
3. .databrickscfg profiles

Profile management

databricks auth profiles          # List all profiles
databricks auth env -p PROD       # Show profile settings
databricks auth token -p PROD     # Show current token
databricks auth describe          # Current auth config details

Use -p <profile> or --profile <profile> on any command. Default profile is DEFAULT.

Global flags

JSON input

Use --json with inline JSON or file reference:

# Inline (Linux/macOS)
databricks jobs create --json '{"name": "my-job", ...}'

# From file
databricks jobs create --json @job-config.json

Filter JSON output with jq:

databricks clusters get <id> | jq -r .cluster_name

Direct REST API access

databricks api get /api/2.0/clusters/list
databricks api post /api/2.0/clusters/create --json '{"cluster_name": "test", ...}'
databricks api post /api/2.0/clusters/edit --json @edit-cluster.json

Supports: get, post, put, patch, delete, head.

Command groups

For detailed subcommands, flags, and examples, see the reference files below.

Compute and runtime

• clusters - Cluster lifecycle (create, start, edit, resize, terminate, delete, pin). See references/compute.md.
• cluster-policies - Cluster configuration rules.
• libraries - Install/uninstall packages on clusters. See references/compute.md.
• instance-pools - Manage cloud instance pools.
• instance-profiles - IAM instance profile administration (AWS).
• policy-families - Available policy templates.

Jobs and pipelines

• jobs - Job lifecycle and run management (create, run-now, submit, cancel, repair, list-runs). See references/jobs-pipelines.md.
• pipelines - Lakeflow / DLT pipeline management and deployment. See references/jobs-pipelines.md.

Workspace and files

• fs - File operations on Unity Catalog volumes and DBFS (cat, cp, ls, mkdir, rm). See references/workspace.md.
• workspace - Notebooks and folders (import, export, list, delete, mkdirs). See references/workspace.md.
• repos - Git repository management. See references/workspace.md.
• secrets - Secret scopes, secrets, and ACLs. See references/workspace.md.
• git-credentials - Personal access token registration for Git.

Unity Catalog

• catalogs - Create, list, update, delete catalogs. See references/unity-catalog.md.
• schemas - Manage schemas within catalogs.
• tables - Table metadata (get, list, delete, exists).
• volumes - File storage with governance. See references/unity-catalog.md.
• grants - Data access authorisation. See references/unity-catalog.md.
• credentials, storage-credentials - Authentication for external storage.
• connections - External data source linkage.
• functions - User-defined function management.
• metastores - Top-level container management.
• registered-models, model-versions - MLflow registry in UC.
• online-tables - Low-latency data access.
• quality-monitors - Data quality metric tracking.

SQL and analytics

• warehouses - SQL warehouse lifecycle (create, start, stop, edit, permissions). See references/sql-analytics.md.
• queries - SQL query CRUD. See references/sql-analytics.md.
• alerts - SQL alert management.
• dashboards, lakeview - Dashboard operations.
• query-history - Query execution history.
• data-sources - Data source listing.

ML and serving

• serving-endpoints - Model endpoint deployment, querying, and AI Gateway. See references/ml-serving.md.
• experiments - MLflow experiment and run management.
• model-registry - Model version and transition management.
• feature-engineering - Databricks Feature Store operations.
• vector-search-endpoints, vector-search-indexes - Embedding search infrastructure.

Identity and access

• auth - Authentication management. See references/identity.md.
• current-user - Authenticated user info.
• users, groups, service-principals - Identity management.
• permissions - Access control for any resource type. See references/identity.md.

Bundles (infrastructure as code)

• bundle - Deploy, run, validate, and manage Databricks Asset Bundles. See references/bundles.md.
• sync - Local-to-workspace directory synchronisation.

Account administration

• account - Multi-workspace account-level management (identity, Unity Catalog, billing, networking, OAuth). See references/account.md.

Utilities

• completion - Shell autocompletion setup.
• labs - Community extension management.
• version - CLI version information.
• configure - Legacy configuration command.

Common patterns

Wait vs no-wait

Long-running operations (cluster start, job run) block by default. Use --no-wait to return immediately, --timeout to set a deadline:

databricks clusters start <id> --no-wait
databricks jobs run-now <job-id> --timeout 30m

Permission management

Most resource types support four permission commands:

databricks <resource> get-permission-levels <id>
databricks <resource> get-permissions <id>
databricks <resource> set-permissions <id> --json @perms.json
databricks <resource> update-permissions <id> --json @perms.json

Pagination

List commands with large result sets support pagination:

databricks jobs list --limit 10 --page-token <token>

Proxy support

Set HTTPS_PROXY environment variable to route requests through a proxy.