On August 25, 2026, Snowflake delivered a cluster of updates that strengthen both metadata management and the day-to-day developer experience. Multi-value tags reached general availability, allowing a single tag on an object or column to hold multiple string values. At the same time, Remote Development with the Snowflake Extension for Visual Studio Code entered public preview, enabling developers to run notebooks, SQL, scripts, terminals, and Cortex Code on Snowflake-managed compute while working from VS Code, Cursor, or Cortex Code Desktop. Complementary networking improvements brought stable egress IP addresses on Azure into preview.
These releases address two persistent enterprise needs: richer, more flexible classification of data assets, and a smoother path for developers to work productively inside the governed AI Data Cloud without constant context-switching.
Multi-Value Tags: Richer Metadata Without Workarounds
Standard tags previously supported only one value per tag per object. That limitation forced awkward patterns—multiple similar tags, concatenated strings, or external metadata stores—whenever an object needed non-mutually-exclusive classifications (for example, multiple data sources or overlapping compliance regimes).
With multi-value tags:
- Create a tag with MULTI_VALUE = TRUE.
- Add or remove individual values using ADD VALUE and DROP VALUE.
- Default limit is 10 values per multi-value tag per object.
- Propagation supports ON_CONFLICT = MERGE (available only for multi-value tags).
- Use SYSTEM$TAG_VALUE_CONTAINS to test whether a specific value is present.
Practical examples
- A table sourced from both CRM and ERP systems can carry DATA_SOURCE = ‘CRM’ and DATA_SOURCE = ‘ERP’.
- A column subject to both GDPR and PCI requirements can hold both compliance values on the same tag.
- Cost-center or project tags can reflect shared ownership without duplication.
Because the tag remains a first-class Snowflake object, existing tag-based masking, data-movement policies, and discovery tools continue to work while gaining the ability to express more realistic classification scenarios.
Remote Development: Snowflake Compute Inside the Local Editor
Remote Development lets teams create a Snowflake-backed development environment and connect to it over Remote – SSH from Visual Studio Code, Cursor, or Cortex Code Desktop. Key capabilities include:
- Execution of Python notebooks and SQL with the same Snowflake Kernel used in Snowsight Workspaces.
- Ability to select and switch among multiple workspaces in the remote environment.
- Full use of Cortex Code (CoCo) inside the remote session.
- Requirement for Snowflake Extension version 1.38 or later; controlled by the account parameter ENABLE_NOTEBOOK_SERVICE_REMOTE_VS_CODE_ACCESS.
Developers keep their preferred local editor and keybindings while the actual compute, data access, and governance remain inside Snowflake. This reduces the friction of moving between local prototypes and production-governed environments and keeps sensitive data and credentials within the platform perimeter.
Related Networking Improvement: Stable Egress IPs on Azure
Also on August 25, stable egress IP addresses entered preview on Azure (already generally available on AWS Commercial). Organizations can retrieve Snowflake egress IP ranges via SYSTEM$GET_SNOWFLAKE_EGRESS_IP_RANGES and add them to external allowlists. On Azure the output includes additional metadata (published date and usage descriptors), making it easier to maintain precise firewall and network-security-perimeter rules for services such as Storage and Key Vault.
How These Updates Improve Discovery, Cost Attribution, and Policy Enforcement
Richer tagging directly benefits:
- Discovery — Multi-value classifications make assets easier to find via multiple legitimate dimensions.
- Cost attribution — Shared ownership or multi-project usage can be expressed accurately.
- Policy enforcement — Tag-based masking, data-movement policies, and access controls can key off any of the values present, supporting more nuanced rules without proliferating tags.
Remote Development complements these governance strengths by keeping developers productive inside the same governed environment that holds the tagged data and enforced policies.
Comparisons to Previous Workflows
Tagging
Previously, teams either accepted single-value limitations, created parallel tags for each possible value, or maintained external metadata catalogs that drifted from Snowflake. Multi-value tags eliminate most of those workarounds while remaining fully native.
Developer Experience
Earlier options included local development with separate connection management, Snowsight Workspaces, or CoCo Desktop. Remote Development unifies the familiar VS Code / Cursor experience with Snowflake-managed compute and the same kernel used in Workspaces, reducing environment drift.
Implications for Metadata Management and Remote Teams
For Data Platform and Governance Teams
- Migrate high-value classification tags that currently use concatenated or parallel-tag patterns to multi-value tags.
- Update tag-propagation and policy logic to take advantage of MERGE conflict handling.
- Document naming and value conventions so multi-value tags remain discoverable and consistent.
For Engineering and Analytics Leaders
- Enable Remote Development for teams that already rely on the Snowflake VS Code extension.
- Standardize on the Snowflake Kernel for notebook work so local remote sessions and Snowsight Workspaces behave identically.
- Combine Remote Development with CoCo for governed, context-aware coding directly against production-like data.
Actionable Insights
- Inventory existing single-value tags that represent non-exclusive categories and plan a controlled migration.
- Pilot Remote Development with a small group of data engineers and analytics engineers; measure reduction in context-switching and environment issues.
- Align network allowlists with the new Azure stable-egress metadata to tighten external connectivity controls.
- Treat multi-value tags as part of the broader metadata strategy that also feeds Horizon Catalog, data-movement policies, and cost reporting.
What This Signals for Platform Usability in Late 2026
The August 25 updates illustrate Snowflake’s dual focus: deepening governance expressiveness while removing friction from the developer loop. Multi-value tags make metadata more faithful to real-world complexity; Remote Development makes the governed platform feel like a natural extension of the tools developers already use.
As organizations scale both AI workloads and distributed engineering teams, platforms that improve classification fidelity and day-to-day usability will see higher adoption of governed patterns. These releases position Snowflake to support that combination through the remainder of 2026.
Conclusion
Snowflake’s general availability of multi-value tags and the public preview of Remote Development with the VS Code extension, both delivered on August 25, 2026, meaningfully advance metadata management and developer experience. Teams can now express non-exclusive classifications natively, enforce policies more precisely, and let developers work in familiar editors while remaining fully inside the AI Data Cloud’s security and governance boundary.
For data platform and engineering leaders, the practical next steps are clear: adopt multi-value tags where single-value limitations have forced workarounds, and evaluate Remote Development as the default path for Snowflake-centric development. Together these capabilities make the platform both more expressive and more usable—exactly what enterprises need as they operationalize governed data and AI at scale.