AI
- AI agent skills for working with databases - DataGrip 2026.2 introduces three agent skills for working with databases: database-connection-management, database-text-to-sql, and database-tools. Skills define reusable procedures for AI agents. They provide step-by-step instructions for completing complex tasks using existing DataGrip features, including MCP tools. Following these instructions helps agents generate higher-quality results with lower token usage.
- Data source management with MCP tools - JetBrains have expanded the list of MCP tools that help AI agents work with databases. This release adds two new tools to the list: create_database_connection and edit_database_connection. The database-connection-management skill includes instructions that help agents use these tools correctly. To have an agent create a new data source, ask it to do so and provide the connection details. The agent will create the data source and run a test connection.
- Other improvements
- AI in DataGrip now remembers query results from previous interactions, allowing you to fetch additional rows without rerunning the query. This also works for statements that return data with RETURNING.
- When an AI agent creates a table, a column, or any other database object, DataGrip automatically introspects the database or schema you're working with. The newly created object will appear in the database explorer right away.
- Previously, when an AI agent couldn't find a referenced database object, it couldn't tell whether the object didn't exist or simply hadn't been introspected. Now, the agent can identify missing objects, suggest similarly named objects in case of typos, and recognize objects that haven't been introspected.
- Queries run by AI agents are now also tracked and logged in the data source query history. To check them, open the data source's query console by pressing Shift+Ctrl+F10 and selecting any query console. In the query console tab, click Browse Query History.
- Previously, AI agents received a generic Not done error when an SQL query failed. Now, the agent receives an actual database message with an SQL error.
- The list_database_connections MCP tool now returns database connection information correctly. Previously, incorrect handling of parameters with default values caused an output schema error.
- The database schema attachment UI in the AI chat has been redesigned.
- Database-specific MCP tools no longer require the AI Assistant plugin.
A smoother start
- Non-modal Welcome screen - The Welcome screen, which appears when you start DataGrip for the first time or after you close all projects, is now non-modal. It appears as a regular IDE tab instead of a modal dialog, so you can continue working in the IDE while using it. This new Welcome to DataGrip tab opens in the automatically created default project. It provides quick actions and useful links to help you get more familiar with the IDE and configure it to suit your needs.
- Interactive onboarding tour - JetBrains have introduced an interactive onboarding tour that guides you through DataGrip's core features. It covers basics such as key actions in the Database Explorer tool window, the data source settings dialog, query execution in a query console, viewing and working with query result sets, and submitting changes to a database.
Command-line interface
- CLI commands for data sources - Power users often need to automate environment setups. To support this, JetBrains have introduced a command-line interface that lets you programmatically manage your database connections and drivers.
Connectivity
- Two-session mode - Before this release, session control was handled through the IDE's default behavior, with the optional Single shared session mode. By default, each new client, such as a query console or data editor, created a separate connection to the database. If Single shared session mode was enabled, DataGrip used one shared connection whenever you worked with a data source. Single shared session mode supports scenarios where you want to share one connection across multiple clients. However, it can be inconvenient because the same session is also used for database introspection - the process of retrieving metadata. DataGrip 2026.2 introduces the new Two sessions: shared and introspection mode, which creates two sessions: one for introspection and one for everything else. This new mode keeps metadata retrieval separate from your working session. If introspection takes longer, query consoles, data editors, and other clients continue using the shared session instead of waiting for it to finish.
- Bundled JDBC drivers - Previously, DataGrip didn't include any JDBC drivers. Each driver you wanted to use had to be downloaded separately when creating a data source. Now, the following JDBC drivers are bundled in DataGrip:
- Microsoft SQL Server 13.2.1.
- MongoDB 1.21.
- MySQL Connector/J 9.5.0.
- PostgreSQL 42.7.3.
- Redis 1.6.
- Role and warehouse switcher Snowflake - Switching between database roles and warehouses is now much faster. If you work with Snowflake, you no longer need to manually execute SET ROLE commands or search through connection settings to change your permissions for the current session. A new drop-down menu in the query console toolbar lets you switch between available roles and warehouses instantly.
Working with data
- SQL expressions in the data editor - You can now set cell values using SQL expressions instead of being limited to literal values. Previously, when editing data in the grid, you could only enter static values or use the context menu to set a cell to NULL or its DEFAULT value. If you wanted to use a database function like NOW() or gen_random_uuid(), or perform a calculation, you had to manually write and execute an UPDATE statement in the console. This update introduces the new SQL Expression mode in the data editor. When editing a cell, you can now right-click it and choose Edit as SQL Expression, or enable SQL Expression mode using the fx button in the value editor.
Database explorer
- Redesigned empty state in the database explorer - The Database Explorer tool window now features a redesigned empty state that offers quick actions for creating data sources in a new project. It also provides PostgreSQL, MySQL, and SQLite demo databases used in the quick start guide, so you can connect to them and get familiar with basic DataGrip features.
- Database colors: v2.0 - Database colors now have a fresh, consistent UI, and custom colors work differently. Previously, the fixed #RRGGBB hex code looked the same across all IDE themes, so it usually looked right in either the light or the dark theme, but not both. Now, you can simply set the hue, and the system handles the shades automatically.
- Schema object grouping by prefix - The database explorer now allows grouping schema objects by prefix in the tree. The following characters are treated as prefix delimiters: _, -, and $. For example, my_table01 and my_table02 are grouped under my_, and their names are displayed as table01 and table02 within the group.
- Deletion of folders and their contents - Previously, you were unable to delete non-empty folders. Now, you can delete folders that contain other folders or data sources. Before deleting a folder, DataGrip will ask for confirmation and display the list of items to be deleted.
Explain Plan menu and actions
- Return of the Explain Plan raw format actions - In the previous release, JetBrains simplified the Explain Plan drop-down by removing the Explain Plan (Raw) and Explain Analyse (Raw) actions. This affected workflows that regularly require these actions. In this release, JetBrains have addressed your feedback by bringing back these actions and making them accessible from the toolbar. There's now a Raw button on the left-hand toolbar of the Query Plan tab. For the plan tree, the button opens the raw execution plan in the Raw inner tab. The plan tree also remains available in the Operation Tree inner tab.
Query files and consoles
- Custom query console names in editor tab headers - Different workflows and tasks require different information to be available at a glance. This is why JetBrains have added the ability to customize how query console names are displayed in editor tab headers - the first place you look when identifying an open file. Previously, only the data source name was displayed next to the file name. Now, you can use variables to add more details, such as the attached data source, database, schema, or search path.
Code editor
- Execution time display near queries - Previously, you could find timestamps for a query execution only in the Services tool window. Now, once DataGrip has executed a query, it shows the query execution time next to the query in the code editor.
- Improved support for ClickHouse and Snowflake SQL - Expanded ClickHouse SQL support with the following additions:
- Query parameter syntax.
- Additional SYSTEM statements.
- The FIRST clause in ALTER TABLE ADD COLUMN statements.
- Other enhancements include:
- Improved support for ORDER BY... WITH FILL STEP.
- Fixes for false syntax errors related to projection indexes.
- For Snowflake, JetBrains have added support for:
- Flow operations.
- DATABASE ROLE syntax.
- HASH functions, including HASH(*).
- INTERSECT, UNION BY NAME, and DISTINCT in SELECT statements.
- Object, session, and user parameters in ALTER USER statements.