Skip to main content

3.0

The 3.0 line covers all 3.0.x releases. Newest at the top.

3.0.3 (20 April 2026)​

This is the initial v3 general-availability release.

What's new​

Release Workflow wizard​

A four-step guided flow takes a change from "I've finished editing" through to "the pipeline has been kicked," handling the version-control commit, deployment-file production, and push / check-in / tag / pipeline trigger in one place. Step labels adapt to the workspace's VCS (Git or TFVC) and task tracker (Jira, Azure Boards, or none), so the same wizard reads correctly across every combination. Every step can still be done manually from the regular tabs; the wizard is an ergonomic path through the common case, not the only path. See The Release Workflow.

MCP server​

DataStar embeds a localhost-only Model Context Protocol server that exposes schema inspection, component search, draft management, deployment generation, and task-tracker integration to any MCP-capable AI client (Claude Desktop, Cline, Kiro, and others). The server is off by default, scoped to the currently-open workspace, and gated behind an auto-generated API key with a separate Allow write access toggle. SQL execution is read-only by default; mutating tools refuse until write access is enabled; connection strings, passwords, and tokens are never exposed through the API. See MCP Overview.

Rebuilt desktop client​

The Windows client has been rebuilt from the ground up. Start-up is significantly faster, the visual design is refreshed throughout, and keyboard navigation has been tightened across every view. The status bar carries new live indicators for the current database connection, version-control state, and basket count.

Deployment Basket as a first-class concept​

What was an implicit staging step in v2 is now an explicit object. The basket has its own mode on the component grid, its own toolbar additions, can be saved to and loaded from a basket file, and feeds directly into a deployment via Create Deployment from Basket. Add and remove components from any grid without leaving the keyboard with Ctrl+B and Ctrl+Shift+B. See The Deployment Basket.

Drafts: review before applying​

A new draft pull mode (Ctrl+D) extracts schema or data into a side-by-side draft file rather than overwriting the workspace. Review, compare, then Apply (Ctrl+Shift+A) or Discard (Ctrl+Z) one row, the current selection, or every draft at once, with progress indicators on the larger sets. See Drafts.

Improvements​

  • Component grid state columns. Workspace, Database, and Versioned states are shown side-by-side on every row, colour-coded for at-a-glance scanning, with hover tooltips for the precise state text.
  • Deployment History tab. The History / Reversal / Summary breakdown is now a full tab rather than a modal dialog, with persistent filters across DataStar restarts.
  • Filter bar on the component grid. Narrow any grid by component name without re-opening the tab.
  • Per-workspace MCP port. Multiple DataStar instances can run side-by-side without colliding on a shared port.
  • Task-tracker parity. Jira and Azure Boards both expose the same set of query, create, update, transition, and comment operations, in the desktop client and through MCP.
  • Component grid context menu. New Compare With submenu for diffing the workspace copy against drafts, the current connection, a different database, HEAD, latest server, remote, or any branch.

Removed in v3​

  • Rally task-tracking integration. Rally customers should move to Jira or Azure Boards, or operate without a task tracker; task IDs can still be entered manually wherever a task is referenced.

Compatibility​

  • Workspaces created in v2 open unchanged in v3; no migration step is required.
  • Deployment files produced in v2 remain consumable by both the v3 desktop client and the DataStar.Tools CLI.
Last updated on