Shared Serverless

Shared Serverless is the current Databricks name, previously Git Folder Serverless (since October 2026).

Git Folder Serverless → Shared Serverless

One serverless compute resource and Python environment shared by every notebook and file in a Git folder or bundle in the editor, with the environment defined by an optional pyproject.toml at the root instead of per notebook.

• Shared compute, separate brains - notebooks on the same resource share the environment and even a web terminal, but a Python variable defined in one notebook is not available in another. 🔗
• Shared, but not with you - only the user who first attaches a notebook or file to the compute can run workloads on it, so the documented way to collaborate is for everyone to clone their own copy of the repository. 🔗
• The preview toggle an admin has to find is not called Shared Serverless at all - it is Project-Scoped Shared Compute, on the Previews page. 🔗

⚠ Limitations: Beta, behind the Project-Scoped Shared Compute preview. Compute sharing is scoped to a serverless usage policy, so notebooks and files in the same Git folder or bundle under different usage policies get separate compute; requires serverless environment version 5 or above; only the user who first attaches can run workloads; Python variables are not shared between notebooks. 🔗

Category

Developer experience

In use from

October 2026🔗

Also known as

Shared Serverless for Git Folders, Shared Serverless for Asset Bundles, Project-Scoped Shared Compute, shared serverless compute for Git folders, pyproject.toml environment

Verified

2026-10-26