How a 20 MB Open‑Source Client Outperforms the 1 GB DataGrip

Last week the author spotted the project dbx at the top of GitHub Trending, noting its 10 k stars and a download size of only about 20 MB. Compared with traditional clients such as DataGrip, DBeaver and Navicat, which are large and resource‑hungry, dbx claims to manage more than 60 database types while remaining lightweight.

Key features

dbx bundles a SQL editor with intelligent completion, virtual scrolling for large result sets, table‑structure management, and import/export capabilities. It also includes an AI‑driven SQL assistant and an MCP Server. Supported back‑ends cover MySQL, PostgreSQL, Redis, MongoDB, ClickHouse, various Chinese databases, vector databases, and even message queues like Kafka and RocketMQ.

Architecture

The tool uses Tauri 2 as the desktop shell and Vue 3 for the UI. The backend is written in Rust , leveraging drivers such as sqlx and tiberius for database connections. Because it does not bundle Chromium, the JRE, or a Python runtime, the binary stays small while still supporting macOS, Windows, Linux, Docker, and a Web version.

Installation

On macOS it can be installed via Homebrew, while Windows users have Scoop and WinGet options. For a quick Web deployment a single Docker command is sufficient:

brew install --cask dbx
# or deploy the Web version
docker run -d --name dbx -p 4224:4224 -v dbx-data:/app/data t8y2/dbx

Offline download

https://github.com/t8y2/dbx/releases

The README also lists additional resources such as ER diagrams, schema diffs, execution plans, field lineage, and instant previews for Parquet, CSV, and JSON files.

Editor experience

What impressed the author most is the editor’s metadata‑aware completion, selectable execution, and history restoration. Result tables can be edited directly, previewed, and exported to CSV, JSON, Markdown, or XLSX. The tool goes beyond merely connecting to a database.

Overall, the author believes dbx is best suited for developers who find traditional clients too heavy yet want AI‑enhanced database workflows. Although the project is still young and some edge‑case UX details need polishing, its 20 MB footprint, support for over 60 databases, and free desktop/Web editions make it worth trying.