DBX: Lightweight Open-Source Database Client

What DBX is

DBX is an open-source database management tool in the same space as DBeaver, Navicat and TablePlus. Its tagline is "100+ databases in 25 MB": a single small native app, with no bundled Java runtime, Python environment or Chromium, that connects to relational databases, NoSQL stores, analytical engines, search engines, vector databases, message queues and configuration services.

The README frames the project around a few complaints with existing tools: DBeaver needs Java, TablePlus is freemium, and most clients stop at databases. DBX ships as a native desktop app for macOS, Windows and Linux, as a self-hosted web version in Docker for team access, as a CLI, and as a separate MCP server so AI coding agents can query databases through connections already configured in DBX.

It is aimed at developers, DBAs and data engineers who work across many different data stores and want one lightweight tool for all of them. The supported list is noticeably strong on databases popular in China (DM, OceanBase, openGauss, GaussDB, KingbaseES, TDengine and others), alongside the usual MySQL, PostgreSQL, SQL Server and Oracle.

How it works

DBX is a Tauri 2 application. The frontend is Vue 3 and TypeScript with shadcn-vue and Tailwind CSS, and the query editor is CodeMirror 6. The backend is Rust, using sqlx, tiberius (SQL Server), redis-rs and the official MongoDB Rust driver for native connections. Because Tauri uses the operating system's webview instead of shipping a browser engine, the install stays around 25 MB.

Not every database has a native Rust driver. For engines such as Snowflake, Trino, Hive, DB2, Neo4j, Cassandra, BigQuery and custom JDBC connections, DBX uses what the README calls agent-based profiles: JDBC agent drivers that live in an agents/ directory built with Gradle and are installed through a driver store. Additional connection types such as S3 browsing, Kubernetes and LDAP arrive as plugins from a built-in store; each plugin package is signature-verified before install and its UI runs sandboxed with its own sidecar process.

The web version runs the same feature set from a Rust backend (dbx-web) behind port 4224. Credentials for connections, plugins, AI providers and SSH tunnels are encrypted before being written to the local dbx.db file. Desktop builds use the platform credential store (macOS Keychain, Windows Credential Manager or Linux Secret Service), while Web and Docker deployments use a managed key at ${DBX_DATA_DIR}/.dbx/secret.key or an explicit key supplied through DBX_SECRET_KEY_FILE or DBX_SECRET_KEY.

Key features

• Broad connectivity: MySQL, PostgreSQL, SQLite, Cloudflare D1, Redis, MongoDB, DuckDB, ClickHouse, SQL Server, Oracle, Elasticsearch, Qdrant, Milvus, Weaviate, TiDB, StarRocks, Redshift, CockroachDB and many more, plus agent-based profiles for Snowflake, BigQuery, Trino and others.
• Query editor: metadata-aware autocomplete, Cmd+Enter execution, formatting, diagnostics, query history, saved snippets and SQL file execution.
• AI SQL assistant: generate, explain, optimize and fix SQL using Claude, OpenAI, local models via Ollama or any OpenAI-compatible endpoint, with safety checks applied before AI-generated SQL runs.
• Data grid: virtual scrolling for large result sets, inline editing with SQL preview before save, DataGrip-style filters, and export as CSV, JSON, Markdown, XLSX or INSERT statements.
• Schema tools: ER diagrams, schema diff across connections, visual explain plans, column-level field lineage and a table structure editor.
• Data operations: CSV and Excel import, cross-database data transfer, full database export, data compare, and drag-and-drop preview of Parquet, CSV and JSON files powered by DuckDB.
• Connection import: bring existing profiles over from DBeaver or Navicat.
• Specialized browsers: Redis (all data types, TTL editing, batch key operations) and MongoDB (document CRUD, Atlas and replica set URLs).
• Middleware consoles: Kafka, RocketMQ, RabbitMQ, Pulsar and MQTT, plus Nacos, Consul, ZooKeeper and etcd.
• Connectivity and safety: SSH tunnels, proxy settings, auto-reconnect, confirmation dialogs for destructive operations and encrypted config export and import.

Getting started

Desktop installs come from GitHub Releases or a package manager. On macOS:

brew install --cask dbx

On Windows with WinGet:

winget install t8y2.dbx

Scoop and Flatpak are also supported. For a self-hosted web version:

# The default keeps the key in the persistent /app/data volume.
docker run -d --pull=always --name dbx -p 4224:4224 \
  -v dbx-data:/app/data \
  t8y2/dbx:latest

Then open http://localhost:4224. Multi-arch images for amd64 and arm64 are published. To expose your DBX connections to an AI coding agent, the MCP server is installed separately:

npx @dbx-app/mcp-server

and registered in .mcp.json:

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

The CLI targets terminal, scripting and agent workflows:

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

To build from source you need Node.js 18 or newer, pnpm and Rust 1.88 or newer; make starts the Tauri development environment and make package produces installers.

Use cases

• One client for a polyglot stack: a team running PostgreSQL, Redis, MongoDB, ClickHouse and Kafka can inspect all of them from one app instead of several.
• Shared browser-based access: the Docker web version gives a team a central database console without installing a desktop client on every machine.
• Letting AI agents query data safely: Claude Code, Cursor or Windsurf can list connections, browse tables and run SQL through the MCP server, with an allowlist and Read only, Data read/write or Full access modes set in DBX Settings.
• Migration and comparison work: schema diff, data compare and cross-database transfer help when moving between engines or verifying environments.
• Quick file inspection: drop a Parquet or CSV file in to preview it through DuckDB without writing a script.
• Leaving DBeaver or Navicat: connection import reduces the setup cost of switching.

How it compares

The README positions DBX against DBeaver (which requires Java) and TablePlus (freemium). Both are established clients; DBeaver in particular has a long history and a very broad JDBC-based driver set, which DBX approaches through its own JDBC agent drivers. Navicat is referenced through the connection import feature. DBX's distinguishing points, as described by the project, are its small native footprint, an Apache-2.0 license, a self-hosted web mode with the same features as the desktop app, first-party MCP and CLI tooling for AI agents, and consoles for message queues and service registries that general-purpose SQL clients usually lack. Maturity and depth per database will vary, so check how well your specific engine is supported before switching.

Things to know before adopting

• Young project: the repository was created in April 2026, so expect rapid change between releases.
• Key management on Docker: the managed secret key must be backed up together with dbx.db. For production, the README recommends supplying the key through a Docker or Kubernetes secret, and explicit keys must not be rotated while encrypted data is in use.
• Credential migration: upgrading from releases that stored credentials in plain text triggers a Data Security Upgrade wizard; the CLI and standalone MCP server will report DATA_MIGRATION_REQUIRED until it has been completed in the desktop or web app.
• MCP is a separate install: installing the desktop app does not install the MCP server. Windows portable builds need DBX_DATA_DIR set in the MCP config.
• Not a sync mechanism: copying dbx.db between devices does not carry the platform keys; use encrypted export and import instead.
• Linux build dependencies: building from source on Ubuntu or Debian requires WebKitGTK and related packages.
• Sponsored README: the README lists commercial sponsors and partners, mostly cloud and AI API providers; the tool itself is free and open source under Apache-2.0.

Project activity

As of October 2026 the repository has roughly 23,800 stars on GitHub, a fast climb for a project created on April 29, 2026. It is written in Rust (with a Vue and TypeScript frontend) and licensed under Apache-2.0. The UI is available in English, Simplified Chinese and Spanish, and the community runs a Discord server alongside QQ, WeChat and Feishu groups. Source code is at github.com/t8y2/dbx and documentation is on dbxio.com.