Remote Development with the Snowflake Extension for Visual Studio Code¶

Remote Development lets you create a Snowflake-backed development environment, connect to it over Remote - SSH, and run Python notebooks, SQL, scripts, and terminals on Snowflake-managed compute. The environment is backed by Snowflake Notebook and includes Python, Jupyter, and Snowflake libraries.

You can use these capabilities in Visual Studio Code, Cursor, or Cortex Code Desktop. All three are built on the same Visual Studio Code extension model, so you install and run the Snowflake Extension for Visual Studio Code the same way in each of them.

To use Remote Development, install version 1.38 or later of the Snowflake Extension for Visual Studio Code and make sure your account meets the requirements in Prerequisites.

Prerequisites¶

Before you start, make sure you have:

Enable and manage access¶

Remote Development is enabled by default through the ENABLE_NOTEBOOK_SERVICE_REMOTE_VS_CODE_ACCESS account parameter. An account administrator can disable it for the account by setting the parameter to false.

When Remote Development is enabled, notebooks in your account can reach the following endpoints through the external access integration that the feature uses:

• update.code.visualstudio.com:443
• vscode.download.prss.microsoft.com:443
• marketplace.visualstudio.com:443
• *.gallerycdn.vsassets.io:443
• download.visualstudio.microsoft.com:443
• marketplace.cursorapi.com:443
• cursor-cdn.com:443
• downloads.cursor.com:443
• anysphere-binaries.s3.us-east-1.amazonaws.com:443
• sfc-repo.snowflakecomputing.com:443
• ai.snowflake.com:443

To turn Remote Development on or off for your own user, open your user settings JSON (Preferences: Open User Settings (JSON) from the Command Palette) and set the value:

{
  "snowflake.remoteDevelopment.enabled": false
}

Set the value to false to hide the Remote Environments panel for your user, or true to show it.

Supported editors¶

Remote Development works in three Snowflake-supported editors that share the Visual Studio Code extension model:

• Visual Studio Code: Install the Snowflake Extension for Visual Studio Code from the Visual Studio Code Marketplace.
• Cursor: Install the Snowflake Extension for Visual Studio Code from the extensions UI the same way you install other compatible extensions.
• Cortex Code Desktop: Install the Snowflake Extension for Visual Studio Code into Cortex Code Desktop. Even though Cortex Code Desktop is itself a Snowflake client, Remote Development ships in the extension today, so you still install the Snowflake Extension for Visual Studio Code to use these capabilities. In Cursor and Cortex Code Desktop, switch to editor mode to follow the instructions in this topic.

Create a remote environment¶

1. In the Remote Environments panel, select Create Remote Development Environment.
2. Enter a Service name. Use a Snowflake-compatible identifier and choose a name that doesn’t conflict with an existing local SSH host alias.
3. Select one or more Workspaces to associate with the remote service. During SSH setup, you’re prompted separately to mount workspaces into the running environment.
4. Optional: select External access integrations if your notebooks need outbound network access, for example to install packages from PyPI.
5. Optional: expand Service settings and adjust the compute and runtime settings.
6. Select Create.

The service appears in the Remote Environments panel. Creation can take a few minutes while Snowflake provisions the service and starts the container.

Service settings¶

Python version and Runtime version appear only when Snowflake Container Runtimes are available for your account.

Manage remote environments¶

The Remote Environments panel lists your remote development services and refreshes every 30 seconds while the panel is visible. You can also refresh it manually.

Use the inline actions next to each service in the panel to manage it:

Suspending a service stops the running remote environment but keeps the service definition so you can resume it later. Compute pool billing and auto-suspend behavior depend on your Snowflake compute pool configuration.

Deleting a service is permanent. The extension also stops any active local proxy for that service.

Connect to a remote environment¶

1. In the Remote Environments panel, find a service with status RUNNING.
2. Select Setup SSH.
3. When prompted, choose one or more workspaces to mount, or press Escape to skip mounting.
4. Your editor opens a new window connected to the remote environment at /root.

You don’t need to create SSH keys or configure port forwarding manually.

If you have many workspaces, there can be a short pause between selecting Setup SSH and seeing the workspace picker while the extension prepares the list. The extension doesn’t show progress during this step, so wait a moment for the picker to appear before selecting Setup SSH again.

Select multiple workspaces¶

You can associate more than one workspace with a remote environment.

• When you create a service or run Setup SSH, the workspace picker supports multiple selection. Choose all the workspaces you want to mount.
• To change the selection for an existing service, use Manage workspaces on remote in the Remote Environments panel. You can add workspaces to, or remove them from, the service.
• In the Explorer view of the remote window, your mounted workspaces appear under Workspaces, already included in the current editor workspace.

Each mounted workspace is linked under:

/root/workspaces/<workspace_name>

If you press Escape at the picker, the extension skips mounting and continues opening the remote window. You can still use the remote environment, but workspace files aren’t linked under /root/workspaces.

If workspace mounting fails, the extension shows a warning and still opens the remote environment. You can retry Setup SSH later after confirming that your role has access to the workspace.

First connection¶

The first connection can take a few minutes because your editor installs extensions on the remote host. If the Extensions view shows Reload Required or Reload Window, reload the remote window before opening notebooks.

If the remote window opens before the Python or Jupyter extension finishes installing, wait for installation to finish and reload the window.

Optional reading: how the SSH connection works¶

You can skip this section if you only need the standard connection workflow.

When you connect, the extension:

1. Starts a local proxy on 127.0.0.1 that forwards SSH traffic to Snowflake over a secure WebSocket connection.
2. Writes a generated Host <service-name> entry to your user SSH config file (for example ~/.ssh/config on macOS and Linux).
3. Installs required extensions on the remote host.
4. Configures the remote Python and Jupyter settings.
5. Opens the remote folder /root in a new editor window.

The generated SSH host entry uses the remote service name. If your SSH config already contains a Host entry with the same name, the extension replaces that entry. Choose service names that don’t collide with SSH hosts you manage manually.

The local proxy uses the Snowflake session token that’s active when you connect. The proxy doesn’t refresh session tokens automatically. If the remote connection becomes unresponsive after a long session, stop the proxy, sign in again if needed, and run Setup SSH again.

Work with notebooks¶

After the remote window opens, create or open a Jupyter notebook:

1. Run Create: New Jupyter Notebook from the Command Palette, or create a file ending in .ipynb.
2. In the action bar of the notebook, select Snowflake: Start Notebook Kernel.
3. In the kernel picker, select Snowflake Kernel (Python + SQL).

This is the same kernel that Snowflake Notebooks use in Snowflake Workspaces. Python and SQL cells run exactly the way they do when you run a notebook in the Snowsight Workspaces UI. Run Python cells and SQL cells in the same notebook without switching kernels.

Run a test Python cell:

from snowflake.snowpark.context import get_active_session

session = get_active_session()
print(session.sql("SELECT CURRENT_VERSION()").collect())

If you mounted workspaces during SSH setup, their files are available under:

/root/workspaces/<workspace_name>

Use mounted workspaces for files you want to keep across remote sessions. Don’t rely on files stored only in local container paths such as /root or /tmp to persist after service lifecycle changes.

Use Cortex Code in a remote environment¶

You can use Cortex Code in a remote environment in two ways:

• In Visual Studio Code or Cursor: Use the Cortex Code support in the Snowflake Extension for Visual Studio Code. For details, see Cortex Code in your code editor.
• In Cortex Code Desktop: Use Cortex Code Desktop as your editor. You still install the Snowflake Extension for Visual Studio Code into Cortex Code Desktop to enable Remote Development, then switch to editor mode to follow this workflow.

Disconnect¶

To end your editor session, close the remote window.

To stop the local SSH tunnel, return to the local editor window and select Stop Proxy for the service. This closes the local proxy and removes the generated Host <service-name> entry from your SSH config.

If you stop or delete a service, the extension also stops any active local proxy for that service.

Troubleshooting¶

Remote Environments panel is not visible¶

Panel is visible but empty¶

You might not have any remote services yet. Select Create Remote Development Environment to create one.

If you expect to see existing services, refresh the panel and confirm that you’re signed in with the expected role and account.

Create fails¶

SSH connection fails¶

Workspace picker is slow to appear¶

If you have many workspaces, the extension can take a moment to prepare the picker after you select Setup SSH or create a service. The extension doesn’t show progress during this step. Wait for the picker to appear rather than selecting the action again.

Notebook kernel doesn’t appear¶

Terminal file listings are slow¶

Mounted workspaces use a remote filesystem. Commands that read many file attributes can be slower than on a local disk. The extension adds an interactive shell alias that disables colorized ls output because color detection can trigger extra metadata calls.

Service is stuck in PENDING or FAILED¶

If a service remains PENDING for several minutes, the compute pool might still be provisioning or might not have available capacity. Refresh the panel and check the service details.

If a service is FAILED, open the Info panel and review the error message. Common causes include invalid compute pools, missing privileges, unavailable runtime images, quota limits, or insufficient compute capacity.

Best practices¶

Use shorter idle timeouts for temporary development sessions to reduce unnecessary compute usage.

Select external access integrations only when your workload needs outbound network access.

Mount a workspace when you need files to persist across sessions or be shared with other Snowflake notebook workflows.

Stop the proxy when you’re done using the remote window so stale SSH entries and local proxy processes don’t remain active.

Reconnect after long breaks. If the Snowflake session used by the local proxy expires, stop the proxy and run Setup SSH again.

FAQ¶

Do I need to manage SSH keys?
No. The extension manages the SSH connection through a local proxy and Snowflake authentication.

Can I open a terminal on the remote environment?
Yes. In the remote window, use Terminal: Create New Terminal.

Can I run SQL cells in notebooks?
Yes. Select the Snowflake Kernel (Python + SQL) kernel. It’s the same kernel that Snowflake Notebooks use in Workspaces, so Python and SQL cells run the same way they do in the Snowsight Workspaces UI.

Can I use Cortex Code in the remote environment?
Yes, in Visual Studio Code, Cursor, and Cortex Code Desktop. See Use Cortex Code in a remote environment.

Can multiple users share one remote service?
Yes, if each user has the required Snowflake privileges to access the notebook service. Remote Development uses SSH access and the Snowflake notebook service privilege model, so multiple authorized users can connect to and operate on the same service in parallel.

What packages are installed?
The selected Snowflake Container Runtime determines the base Python version and preinstalled packages. For runtime details, see Snowflake Container Runtime for ML.

Can I install additional packages?
Yes, if the environment has outbound network access through an external access integration or another approved package source.