DataGrip 2026.3: Agentic Development With JetBrains Air, Cross-Database Queries, DuckDB Support, and Major UX Improvements
The JetBrains Blog · 软件开发
DataGrip 2026.3, the third major update of the year, is here! Let’s take a look at what’s inside. AI JetBrains Air is now available by default in DataGrip. You can run SQL, explore data, and reference database objects directly in chats with AI agents. SQL snippets use your IDE’s syntax highlighting and formatting settings. You can run them against a selected data source using the Run Query button right in the chat. After execution, you can navigate to object definitions and preview them in the Quick Definition popup. You can view CSV format data in the chat’s data editor with sorting and search options. To reference database objects in your prompts, use @dbObject notation with completion to help you find the object you need. Object references in the chat also support Quick Definition . UI and UX improvements First, we’ve moved query results from Services to a new Database Output tool window. It opens automatically at the bottom of the IDE when you run a query, a run configuration, or an SQL file. By default, its contents follow the active editor tab. The header combines the client selector with the Output , Query Plan , and Result tabs. The session client tree is now in a dropdown on the right side of the toolbar, while session management remains in the Services tool window. Second, we’ve updated the tool window bar in the window header and the default tool window shortcuts. The bar now shows icons for Database Explorer , Database Output , Files , and Agent Sessions by default. For the default tool window shortcuts, Agent Sessions now uses Cmd+2 (macOS) / Alt+2 (Windows/Linux), previously assigned to Files . Files now uses Cmd+4 (macOS) / Alt+4 (Windows/Linux), previously assigned to Run . Run no longer has a default shortcut. Connectivity You can now query data across different data sources, even when they use different DBMSs, and view the results directly without copying the data beforehand. To make this possible, we’ve introduced a new Workspace data source, base