Connect from your IDE

This page summarizes features of developer tools that enable you to connect to Databricks from your IDE.

Compare IDE tools

The following table compares tools for connecting to Databricks from your IDE.

Tool	Use cases
SSH tunnel	Connect your local IDE to classic, serverless, or serverless GPU compute over SSH. Run code in the same environment as your cluster or serverless compute. For additional feature details, see IDE connected with SSH.
Databricks Sandbox	Connect your local IDE to lightweight Databricks serverless compute over SSH. Default environment includes common coding agent tools and support for Omnigent. For additional feature details, see IDE connected with SSH.
IDE extension	Connect Visual Studio Code or Cursor to a Databricks workspace with a few clicks. Define, deploy, and run Declarative Automation Bundles from your IDE. Run local Python files on Databricks compute, or run files and notebooks as Lakeflow Jobs.
Databricks Connect	Connect any IDE, notebook server, or custom application to Databricks compute. Write code with Spark APIs and run it remotely on Databricks compute instead of a local Spark session.

Tool	Use cases
SSH tunnel	Connect your local IDE to classic, serverless, or serverless GPU compute over SSH. Run code in the same environment as your cluster or serverless compute. For additional feature details, see IDE connected with SSH.
Databricks Sandbox	Connect your local IDE to lightweight Databricks serverless compute over SSH. Default environment includes common coding agent tools and support for Omnigent. For additional feature details, see IDE connected with SSH.
IDE extension	Connect Visual Studio Code or Cursor to a Databricks workspace with a few clicks. Define, deploy, and run Declarative Automation Bundles from your IDE. Run local Python files on Databricks compute, or run files and notebooks as Lakeflow Jobs.
Databricks Connect	Connect any IDE, notebook server, or custom application to Databricks compute. Write code with Spark APIs and run it remotely on Databricks compute instead of a local Spark session.

IDE connected with SSH

Both the SSH tunnel and the Databricks Sandbox connect to Databricks compute in your workspace with SSH (Secure Shell). The following table compares use cases and technical differences of the two features.

	Databricks SSH tunnel	Databricks Sandbox
Primary use case	IDE-driven development for data engineering, data science, and machine learning workloads on Databricks compute.	Run agent loops and safely execute untrusted, arbitrary non-Spark and Databricks Runtime code.
Where compute runs	Either classic or serverless data plane in the Databricks workspace	Serverless data plane in the Databricks workspace
Compute type	Serverless compute Dedicated single-user clusters AI Runtime	Special lightweight serverless compute
Start time	Similar to existing compute startup times	Seconds
Identity propagation	Automatically authenticated to the user who initiates the SSH session	Automatically authenticated to the user who initiates the SSH session
Data persistence	You have a direct connection to the Databricks workspace. Data in workspace files and Unity Catalog is available from the IDE UI, REST API, and Databricks CLI.	Data is available in /home/sandbox-agent/, which persists when the sandbox is stopped. Data in workspace files and Unity Catalog is available with the REST API and Databricks CLI.
Default environment	Same as your cluster or serverless environment. Manage dependencies on the cluster, leverage workspace base environments for serverless and AI Runtime, or install dependencies during project runtime.	Default environment includes common coding agent tools and support for Omnigent.
SEG (egress controls)	Inherits from workspace	Not yet available
Ingress controls (PrivateLink, IP ACL)	Available	Available

	Databricks SSH tunnel	Databricks Sandbox
Primary use case	IDE-driven development for data engineering, data science, and machine learning workloads on Databricks compute.	Run agent loops and safely execute untrusted, arbitrary non-Spark and Databricks Runtime code.
Where compute runs	Either classic or serverless data plane in the Databricks workspace	Serverless data plane in the Databricks workspace
Compute type	Serverless compute Dedicated single-user clusters AI Runtime	Special lightweight serverless compute
Start time	Similar to existing compute startup times	Seconds
Identity propagation	Automatically authenticated to the user who initiates the SSH session	Automatically authenticated to the user who initiates the SSH session
Data persistence	You have a direct connection to the Databricks workspace. Data in workspace files and Unity Catalog is available from the IDE UI, REST API, and Databricks CLI.	Data is available in /home/sandbox-agent/, which persists when the sandbox is stopped. Data in workspace files and Unity Catalog is available with the REST API and Databricks CLI.
Default environment	Same as your cluster or serverless environment. Manage dependencies on the cluster, leverage workspace base environments for serverless and AI Runtime, or install dependencies during project runtime.	Default environment includes common coding agent tools and support for Omnigent.
SEG (egress controls)	Inherits from workspace	Not yet available
Ingress controls (PrivateLink, IP ACL)	Available	Available