dbx: a 25 MB database client for 100+ databases

TL;DR: dbx is a database client written in Rust that weighs 25 MB and claims support for more than 100 database engines. It ships four ways to reach the same thing: a desktop app, a CLI, a Docker image, and an MCP server. At 22,000 stars it is already one of the larger Rust projects in the tooling space.

What it is

The description names eight engines outright: MySQL, PostgreSQL, SQLite, Redis, MongoDB, DuckDB, SQL Server, and Dameng. Dameng is the interesting name in that list, because it is a Chinese enterprise relational database that most Western client tools ignore. Its presence suggests the project is not scoped to a single market.

The size claim is the headline. A 25 MB client that covers that many engines is a different proposition from a 300 MB Electron app that covers eight. If you have ever installed a database GUI on a laptop with limited disk, or pulled one into a container image, you already know why that number matters.

The repository is bilingual. Every description element ships in English and Chinese, which is more than a translation courtesy. It tells you the intended user base includes people for whom a Chinese-language database client is the difference between a usable tool and an unusable one.

Four surfaces, one tool

What makes dbx worth a look is not the database coverage, it is how many ways the project lets you in.

Surface	What it is for
Desktop app	Daily interactive work on a local machine
CLI	Scripting, containers, CI, anything without a display
Docker	One container, mount credentials, reach it over the network
MCP server	Let a coding agent query and inspect your databases

The MCP server is the piece that changes the conversation. If your agent has a connection to your production database, “show me the row counts on the orders table” stops being a question you paste into a SQL editor and becomes a tool call. That is genuinely useful for exploration and genuinely risky for writes, which is why the boundary between read and write access matters more here than the rest of the tool.

The CLI and Docker images are what make dbx usable in places a desktop app cannot reach. A laptop install is fine. A container in a bastion host, or a CLI binary in a build script, is where database clients actually earn their keep.

Written in Rust

The primary language is Rust, and the license is Apache-2.0. Both of those are load-bearing facts here rather than trivia.

Rust is the usual answer when someone wants a single static binary that starts instantly and has no runtime dependency to install. That directly explains the 25 MB claim. A Node or Python client would need a language runtime present, and a packaged Electron client would carry a browser with it.

Apache-2.0 is permissive, which means you can vendor dbx into an internal tooling bundle, modify it, and redistribute it without the licensing negotiation that copyleft licenses would force. For a tool that touches credentials and production databases, having the option to self-build and pin an exact version is worth more than it sounds.

Reading the trajectory

The repository was created at the end of April 2026 and now sits at 22,108 stars and 2,091 forks. A fork ratio just under 10 percent on a project this new is high, and it lines up with the licensing and the Rust build story. People are not only trying the tool, they are taking the source.

Activity is current. The last push landed on the same day this was written, which for a client with a hundred-odd backends is the number to watch, because backends break. A driver that stops compiling against a new server release is the failure mode, and daily commits suggest someone is watching for it.

How to try it

The site is at dbxio.com. Start with a database you do not mind breaking, connect over a read-only account, and see whether the engine you actually use is on the list. If it is, the next thing worth testing is the MCP server, because that is the surface with the most leverage and the least obvious failure modes.

When it is not what you need

If you already have a database client you like, the switching cost is the driver coverage, not the interface. Fifty engines you never use are worth less than the ten you use every day, so verify your specific engine first.

It is also the wrong shape for a team that has already standardised on a heavier GUI and written tooling against it. The MCP server is the piece most likely to survive a trial even if the desktop app does not.

Source and Accuracy Notes

• Project page: https://github.com/t8y2/dbx
• Source repository: t8y2/dbx (verified via GitHub API html_url)
• License: Apache-2.0 (verified via GitHub API license.spdx_id)
• Stars at time of writing: 22,108
• Project site: https://dbxio.com
• Engine counts, the 25 MB figure, and MCP server support are taken from the repository description as published by the maintainers. Individual engine behaviour was not independently tested.