DBX v0.6.51: New SQL Formatter, :dbx Vim Commands, and Dozens of Improvements

Image: GitHub — the t8y2/dbx repository

DBX — a lightweight open-source database client (around 25 MB for 100+ databases) — has just shipped its latest version: v0.6.51, released on October 21, 2026. Even though it's still 0.x, DBX is one of the most productive open-source projects out right now: v0.6.51 is the 365th release on GitHub, with an essentially daily release rhythm.

This release brings 8 new features, around 24 improvements, and more than 35 fixes — almost all of them community contributions. If you're new to it: DBX is a cross-platform database client (Rust + Tauri 2, Apache-2.0 license, 27k+ stars) supporting MySQL, PostgreSQL, SQLite, Redis, MongoDB, ClickHouse, DuckDB, Oracle, Dameng, various vector databases, message queues, and much more — complete with a built-in AI assistant and MCP server.

Now, let's look at what's new in v0.6.51.

TL;DR: DBX for the Uninitiated

Before diving into the release details, a bit of context for anyone hearing this name for the first time. DBX is something like a spiritual successor to DBeaver/Navicat, but with a different philosophy: lightweight first, feature-complete later. The desktop installer is only about 25 MB — far below DBeaver, which needs hundreds of MB (or Electron, which bundles its own Chromium; DBX uses Tauri 2, which calls the system WebView). The result: the app is fast, memory-friendly, and works offline.

But lightweight doesn't mean bare-bones. Here are the core features DBX already has:

• SQL editor (CodeMirror 6) with metadata-aware autocomplete, formatting, history, and execution of selected SQL.
• Virtualized data grid with inline editing, WHERE/ORDER BY controls, SQL preview, and export (CSV, Excel, SQL, full export).
• Schema browser & ER diagram — navigate tables, columns, indexes, foreign keys, triggers, plus field lineage.
• Schema diff & migration — compare structures, import tables, export databases, execute SQL files, cross-engine transfers.
• SSH tunnel, encrypted config export, and guards for destructive actions (production-safe).
• AI assistant that can write, explain, optimize, and fix SQL — plus an MCP server so Cursor/Claude Code can query your database directly.
• Support for 100+ databases: from mainstream MySQL/PostgreSQL/SQLite and Redis/MongoDB to vector DBs (Qdrant, Milvus, Weaviate, ChromaDB), time-series (InfluxDB, QuestDB, TDengine), analytics (ClickHouse, DuckDB, Doris), message queues (Kafka, RabbitMQ, Pulsar), and China-market-specific ones (Dameng, GaussDB, KingbaseES, OceanBase, PolarDB).

The project itself runs in three modes: desktop (macOS, Windows, Linux), Docker for browser-based access (self-hosted), and a CLI for terminal/automation. All from the same project, under the Apache-2.0 license.

New Features in v0.6.51

1. SQL Formatter with Configurable Rules

Contributed by @efeire (PR #12772). SQL formatting is now much more flexible: line breaks for independent lists, logical line breaks in WHERE/JOIN ON, blank lines between code blocks, even line width — everything can be tuned to match your team's style. The settings panel even has a live preview, so you see the result before saving. Off by default, so anyone who likes the old formatting style won't get a surprise.

2. "Execute All SQL" Button + Alt+X Shortcut

Contributed by @lxk955 (PR #12778). The editor toolbar now has an Execute All SQL button to run the entire editor contents at once, plus an Alt+X shortcut for keyboard people. Small, but it makes the migration-script workflow way faster.

3. Vim Mode: :dbx Commands

Contributed by @youmin1017 (PR #13044). For the true Vim faithful: Vim mode in DBX now has a set of :dbx commands that run DBX operations straight from the editor's command line — and more importantly, they can be mapped in vimrc. Want :w to save and run the query? Set it up yourself.

4. "Two-Row Grouping" Tab Layout

Contributed by @yaoxinghuo (PR #12905). When your horizontal tabs have piled up (happens to everyone with lots of open connections), there's a new layout option: the first row shows groups, the second row shows the tabs in the active group. Much tidier than an endless strip of tabs.

5. Refresh Redis Prefixes Without Reloading the Whole Tree

Contributed by @caichangqing1120 (PR #13050). In the Redis panel, key-prefix directory nodes can now be refreshed individually without reloading the entire database tree. For anyone working with Redis instances holding thousands of keys, this is a significant time-saver.

6. AI Assistant Reads Database Docs

Contributed by @Abeautifulsnow (PR #12705). This is the "smartest" feature in the release. When the AI assistant assembles the database structure context for the model, it now reads the Database Docs notes on that connection. That means: the business meaning of tables and columns ("trx_amt is the transaction amount in rupiah, not USD") can be understood by the AI without being written into database comments. Docs stay in DBX, the database stays clean.

7. Plugin Store with Install Counts

Contributed by @peakxy (PR #13083). The plugin store now shows the install count for each plugin and supports sorting by install count — a practical way to see which plugins the community actually trusts.

8. Keyboard Navigation in the SQL File Panel

Contributed by @eryajf (PR #13076). The SQL file panel now supports full keyboard navigation: arrow keys to select, Enter to open, F2 to rename — plus a single-click or double-click setting for opening files.

Improvements Worth Noting

The improvements section of this release is long, but here's what matters most for daily use:

Execution & query results:

• Statements without a result set (INSERT, UPDATE, DDL) now automatically show a Summary tab with execution status — no more going blank with no news.
• Batch execution of many statements automatically opens a summary view with per-statement progress.
• Running SQL without selecting a connection now pops up a connection picker dialog instead of a confusing query error.
• Table results have a "clear remembered column width" action to go back to auto-width.
• Edited cells in paginated tables no longer wipe out Ctrl+F search results.

Editor & autocomplete:

• A new "enable sequence completion" toggle to turn off sequence object & NEXTVAL/CURRVAL suggestions.
• MySQL completion now suggests AES_ENCRYPT / AES_DECRYPT functions.
• Copy DML now includes fully qualified database names, so it doesn't land in the wrong place when pasted into another session.
• Hover object navigation (Ctrl/Cmd+click) now covers stored procedures and functions.

Connections & navigation:

• Ctrl+P quick open has a connection scope dropdown to search objects within a single connection.
• SQL Server connections accept the port directly in the host field (host,port, host:port, or a tcp: prefix).
• The data generator has a Unix timestamp option (seconds & milliseconds).
• The production environment marker is now a neat red shield at the end of the line, and the table toolbar uses a Connection / Database › Table breadcrumb.

Fixes: Highlights

More than 35 bugs fixed. The most interesting ones:

Cross-platform & UI:

• The visual bug on Linux GNOME X11 (transparent window, only the mouse area redrawing) is finally fixed.
• Hiding the "check update" button now genuinely stops silent update checks in the background.
• Restoring SQL from history now opens it in a new query tab — no longer overwriting the editor you're working in.
• Font/density changes no longer wipe manually set column widths.

Autocomplete & editor accuracy:

• Completion is no longer empty when the cursor is inside a quoted identifier.
• The SQL formatter rewrite now indents BEGIN/END, IF, and LOOP blocks in stored procedure bodies correctly.
• Unqualified object names are now resolved against the active schema — views are no longer mistaken for tables.

Database-specific fixes:

• Oracle: an input dialog for replacement variables (&name / &&name); stream export now uses ROWNUM pagination; DDL transfer no longer carries over source tablespace attributes (fixes error ORA-01950).
• MySQL: connection pool timeout now follows the "connection timeout" setting; schema diff ADD COLUMN preserves charset & collation; auto-increment on a unique key no longer produces illegal DDL.
• PostgreSQL: fixes for large values being truncated in the preview before saving; schema diff no longer false-positives on stored procedures; TBase numeric columns display correctly (no NULLs).
• Kingbase, GBase 8s, OceanBase (Oracle mode), Nacos 2, MQTT — each gets its own compatibility fixes.

This release also brings an important fix for cross-engine data transfer (MySQL bigint/tinyint → Oracle mapping, and others) and backup retention policy, which now cleans up emptied run directories.

Ecosystem: MCP Server, CLI, and Agent Skill

What keeps DBX relevant in the age of AI coding agents (unchanged in v0.6.51, but worth remembering):

MCP Server. DBX has an official MCP server you can plug into Claude Code, Cursor, Windsurf, or any MCP-compatible agent:

{
  "mcpServers": {
    "dbx": { "command": "npx", "args": ["-y", "@dbx-app/mcp-server"] }
  }
}

Its permission mode is managed neatly from DBX Settings → MCP with three levels: Read only, Data read/write, and Full access. For web/Docker deployments, point the MCP server at the web backend API.

CLI + Agent Skill. For terminal and agent workflows:

npm install -g @dbx-app/cli
# or via Homebrew
brew tap t8y2/tap && brew install dbx-cli
 
dbx agent setup          # install DBX Agent Skill offline
dbx connections list --json
dbx query local "select 1" --json

The dbx agent setup command installs the official DBX Agent Skill in ~/.agents/skills/dbx, so any agent that can run a shell knows how to use the DBX CLI safely. There are also native binaries (no Node.js needed) for macOS, Linux, and Windows — great for offline environments or servers.

How to Install & Update

# macOS (Homebrew)
brew install --cask dbx
brew upgrade --cask dbx        # update to v0.6.51
 
# Windows (Scoop)
scoop bucket add dbx https://github.com/t8y2/scoop-bucket
scoop install dbx
 
# Windows (WinGet)
winget install t8y2.dbx
 
# Linux (Flatpak)
flatpak remote-add --if-not-exists flatpark https://dl.flatpark.org/flatpark.flatpakrepo
flatpak install flatpark com.dbxio.dbx
flatpak update                 # routine update
 
# Docker (web mode)
docker pull dbxio.com/dbx:v0.6.51

The desktop installer is available for macOS (Apple Silicon & Intel), Windows 10/11, Windows 7/Server 2012 R2, Linux x64/ARM64, plus browser packages for Linux distros like Kylin and UOS. If GitHub downloads are slow for you (especially from China), there's also a CNB mirror and the Docker image docker.cnb.cool/dbxio.com/dbx:v0.6.51.

Quick FAQ

Is updating to v0.6.51 safe for my configuration and connections? Yes. DBX stores app data (including the connection list) separately from its binary, and 0.6.x releases follow a regular patch/minor cycle. Still, the usual advice: back up or export your encrypted configuration first if you keep a lot of important connections.

I use DBeaver/Navicat. Should I switch? Depends on your needs. If you need specific enterprise features (advanced DBA tooling, for example), DBeaver/Navicat may still go deeper. But for daily use — querying, editing data, browsing schemas, light migrations — DBX offers a lighter experience that's free. And if you're starting to fold AI into your database workflow, DBX's built-in MCP server is a plus the old clients don't have.

Will the new SQL formatter rewrite all my SQL the moment I enable it? No. These new formatting rules are off by default and only apply when you format — the settings panel has a live preview before you apply anything.

What's the difference between desktop and Docker mode? Desktop runs locally on your machine (fastest, best for daily work). Docker runs the web version, accessible through a browser — good for team servers or multi-device access. The CLI is separate, for terminals and automation.

Why is the version still 0.x when the features feel mature? DBX does stick with 0.6.x numbering even though it's had hundreds of releases and wide usage. Don't be fooled by the number — the activity level (365 releases, thousands of PRs) shows a project under very active development.

Does the AI assistant need a paid API key? The DBX AI assistant can be configured with your model provider of choice (including local ones). For agent workflows, the MCP server works with Claude Code, Cursor, or whatever agent you already use.

Conclusion

DBX v0.6.51 might look like a routine release, but look at what's inside: a configurable SQL formatter, :dbx commands for Vim users, AI that now understands the business meaning of columns through Database Docs, plus dozens of community improvements and fixes — this is a release that makes daily use smoother on almost every front.

If you've never tried DBX and have been stuck between expensive Navicat and heavyweight DBeaver, this one is worth a shot. And if you're already a user, run brew upgrade --cask dbx (or flatpak update) and enjoy the fixes.

Happy querying! 🐳

References

• DBX v0.6.51 Release Notes (GitHub)
• t8y2/dbx — GitHub repository
• DBX — official site & documentation
• DBX Changelog