July 2026
- ImprovedMCP ServerJul 18
Altimate Code chat shows what it's doing, live
Altimate Code now makes agent work visible as it happens. When the chat needs time to inspect project files, call tools, or gather context, users can see live progress instead of waiting without feedback. This makes the experience feel faster, clearer, and more reliable, especially for longer-running workflows. Tool steps are also shown with friendly labels and source badges, so users can quickly understand what the agent is doing and which type of tool it is using.
- ImprovedMCP ServerJul 18
Altimate Code installs on Windows without Node
Installing or updating Altimate Code on Windows no longer requires Node or npm. The Install panel and the background updater now use a PowerShell installer that drops the Altimate Code binary straight into place, for both first-time installs and later updates. If a locked-down Windows machine ever made installing via npm painful, that step is gone.
- New
Power user for dbtJul 18Explain and profile queries with Altimate Code
Altimate Code now brings query explanation and performance profiling directly into the Query Results panel. After running a query, users can ask Altimate Code to explain the compiled SQL or profile the query with the full context already loaded. This makes it easier for analytics teams to understand results, investigate slow queries, and move from question to next step without leaving dbt Power User.
- ImprovedMCP ServerJul 5
Altimate Code auto-installs on first chat open
Opening the chat panel for the first time now installs Altimate Code automatically in the background instead of sending you to a manual install screen. You'll see a short "Preparing your Altimate Code experience" notification while it sets up, with the old manual install flow kept as a fallback if auto-install fails. Prefer to do it yourself? Turn off the new auto-install setting in your editor preferences.
- Improved
Snowflake App
Databricks AppJul 5Better error messaging when no team is selected in Studio
Asking Studio any question without a team selected now shows a clear message prompting you to pick a team, instead of a generic error with no obvious next step.
- ImprovedMCP ServerJul 5
Datamate MCP server defaults to stdio transport
The datamate MCP server now uses the stdio transport by default, so external MCP clients like Claude Code, Claude Desktop, and Cursor connect out of the box. Previously you had to manually switch the transport setting to get this; now it's the default, and you can still switch back if you prefer the old behavior.
- ImprovedSnowflake AppDatabricks AppJul 5
Enterprise AI token usage is now metered and visible
Enterprise and premium plans previously bypassed usage metering entirely, so there was no way to see how many AI tokens your team had used. Usage is now tracked against your provisioned grant, the same way every other plan already works, and shows up in the same token usage view — so you can see exactly what's been used and what's left.
- ImprovedDatabricks AppJul 5
Filter AI/ML opportunities by endpoint name
You can now filter Databricks AI/ML cost-saving opportunities by endpoint name. Type or paste part of a name to jump straight to the model-serving, vector-search, or AI Gateway endpoint you're looking for. The filter also works alongside the existing Type filter, and accepts letters, numbers, and symbols.
- ImprovedDatabricks AppJul 5
Fixed Blank Runtime Engine and Photon fields
The Runtime Engine and the Photon Enabled fields always showed a blank dash no matter how the cluster was actually configured. Both now populate correctly from the cluster's runtime version, so you can see at a glance whether Photon is turned on.
- ImprovedDatabricks AppJul 5
Fixed Workspace-scoped visibility for model-serving endpoint opportunities
Fixed: Access scoping for model-serving endpoint opportunities Model-serving endpoint opportunities are now properly scoped by workspace access, matching the behavior of cluster, job, and warehouse opportunities.
- BetaDatabricks AppJul 5
New (Beta): Auto Tune comes to all-purpose clusters
Auto Tune now supports Databricks all-purpose clusters, with a dedicated tab matching the existing Jobs experience. It includes: - Cost-savings chart to track the savings that Auto Tune is helping you realize - Audit history of all Auto Tune activity on the cluster - Toggles for Worker Sizing and Auto Termination recommendations You're in control of what Auto Tune can do by enabling just the recommendation types you want. Once enabled, Auto Tune operates autonomously, applying optimizations with SLA-safe rollbacks if anything regresses. Opening any recommendation shows a before/after diagram of the cluster's current and proposed configuration, so you always have full visibility into what's changing.
- ImprovedSnowflake AppJul 5
New Site name column on Tableau connection sync history
For Tableau connections spanning multiple sites, the Sync History table now includes a Site Name column showing which site each sync run belongs to. Previously, sync events weren't attributed to any site.
June 2026
- NewPower user for dbtJun 20
ERD overlay in the lineage graph
The Power user for dbt lineage graph now makes it easier to understand how your data models connect. With the new ERD overlay, you can view table relationships directly in the lineage graph, helping you trace dependencies, review foreign-key structure, and understand model relationships without switching to a separate ERD tool.
- NewPower user for dbtJun 20
Generate dbt unit tests from the documentation editor
Generate dbt unit tests faster from the documentation editor. The Unit Tests row now makes it easy to create dbt 1.8+ manifest unit tests for the model you are reviewing, with Altimate Code Chat opening directly in context to help draft a new test in one click.
- ImprovedPower user for dbtJun 6
Fix with Altimate Code now meets you at every dbt failure
Fixing dbt failures is now faster and more seamless inside Power User for dbt. When a run, build, compile, parse, or SQL validation error appears, users can launch Altimate Code directly from the failure with one click. Altimate Code opens with the relevant command, resource, and error context already included, so users can move from failure to guided troubleshooting without copying error messages, switching tools, or rebuilding context manually.