Releasesby GlyphSandboxes for the Hugging Face MCP Server
Sandboxes give an assistant secure execution environments attached to buckets and repositories for dataset analysis, model training, and Space creation.
Also shipped Jul 22, 2026Hugging Face in July 2026 →
External OAuth and configurable scopes for MCP servers (General availability)
Two new hierarchical parameters let you bind MCP servers to your organization’s existing identity provider (IdP) and control which OAuth scopes MCP clients discover. This enables MCP clients such as Claude, Cursor, and ChatGPT to authenticate against your corporate IdP instead of Snowflake OAuth.
Variant and Variant shredding are now generally available
Variant is now generally available . Variant is a flexible data type for semi-structured data that delivers structured-like query performance through Variant shredding, which is also generally available.
Transactions are now generally available
Transactions that write to Unity Catalog managed Delta tables are now generally available. Group multiple SQL statements across one or more tables into a single atomic transaction using BEGIN ATOMIC ... END; or BEGIN TRANSACTION; ... COMMIT; syntax. All changes succeed together or roll back together, ensuring data consistency across your operations.
Email notifications for expiring service principal tokens are now generally available
Databricks now automatically sends an email notification when a service principal token is set to expire within seven days. Because service principals don't have their own mailboxes, Databricks sends these notifications to workspace admins.
Role-based access control (RBAC) is in Public Preview
Role-based access control (RBAC) is now in Public Preview. RBAC lets users assume a role in Databricks using only that role's permissions for the duration of the session instead of their own accumulated permissions. This enables exclusive access.
Governance Hub is in Beta
Governance Hub is in Beta . Governance Hub is a centralized, account-level UI in the account console for monitoring governance across your Databricks account. It provides a unified view of data health and coverage, AI usage and spend, and cost drivers, with prioritized insights and recommendations. All views respect existing permissions.
More from Hugging Face HubEvery Hugging Face Hub release →
Hugging Face MCP Server adds hf_fs tool
The Hugging Face MCP Server's new hf_fs tool provides a single interface to repositories, storage, documentation, papers and more, with search, in just over 1,000 tokens.
Egress metrics for users and organizations
Users see egress usage in the dashboard and organizations get a per-user egress breakdown. It currently covers only traffic routed through the Hugging Face CDN.
Build Spaces with AI agents
The new Space creation page has an option to build with an AI agent: copy the generated command into your agent to build and iterate on a Space for a model, paper, or local folder.
Fine-grained access token presets
Fine-grained tokens can be created from presets (Read-Only, Inference, Write, CI/CD, Full Access), and preset URLs link straight to a pre-configured token.
Filter HF Jobs by label
Jobs can be filtered by label, with most-used labels shown as chips with job counts and a free-form key=value input, on user and organization jobs pages.
Filter Jobs by Label
Users can now filter Jobs by label on both user and organization jobs pages. Most-used labels appear as clickable chips with job counts, and a free-form key=value input allows filtering by any label.
Granular feature access per resource group
Feature access can be controlled per resource group instead of only by organization role, for example leaving Jobs open, restricting Inference Endpoints to admins, and limiting blog publishing.
Granular Feature Access
Organizations can now control feature access per resource group rather than across the whole organization. This allows restricting features like Inference Endpoints to admins while leaving Jobs open to everyone.