DBX: A 25 MB Database Client for 90+ Databases — With Your Own AI on Top

1.What DBX Is: 25 MB, 90+ Databases, Open Source0:00

DBX is a database client that weighs in at 25 MB and connects to 90+ databases — PostgreSQL, MySQL, SQLite, Redis, MongoDB, DuckDB, SQL Server, ClickHouse, Dameng, and more. Joe Maddalone's pitch is blunt: "it's probably going to replace any other clients you might be working with now."

It's open source (github.com/t8y2/dbx, Rust, Apache-2.0, ~21K stars), which the video notes means you can "peruse the code or fork it or whatever you want to do." The whole thing is one small binary that ships a desktop app, a Docker image, a CLI, an MCP server, and a web API.

2.Connect & Query: Autocomplete and Command-Enter0:30

The workflow starts with New Connection, where you pick from a long list of database types. The demo uses a plain PostgreSQL connection — name, IP, user, password — then a test connection that confirms before you save. Once connected, the left panel lists your tables and you can drill into any database.

Querying is immediate: a new query window gives you autocomplete as you type, and Cmd/Ctrl-Enter runs it and drops the results below. That's the whole core loop — connect, type, run, read — with none of the ceremony of the big clients.

3.The AI Layer: Point It at Your Own Endpoint1:20

The feature that makes DBX stand out is the built-in AI, and the detail that matters is that it's provider-agnostic. Under Settings → AI → Add Config, you name a config, mark it OpenAI-compatible, drop in an API key, and point it at a URL.

The config has a test button that pings the endpoint before you apply it, so you know the AI is actually reachable before you try to use it.

4.AI in Action: Explain Schema, Generate SQL1:48

With the AI configured, the demo is two taps: hit the little robot button, pick your connection, and type "Explain the database structure to me." It returns a plain-language rundown of your tables and relationships — the fastest onboarding there is for a schema you didn't write.

Then the useful part: you can build complex queries with AI and apply them directly to the database. Natural language in, SQL out, executed in place — no copying between a chatbot and your client. For a developer who knows SQL but is tired of writing boilerplate joins, that's the killer feature.

5.Beyond the GUI: CLI, MCP, Docker, Web2:16

The video closes by flagging that the desktop app is "just the surface." DBX also ships a CLI, an MCP server, a web-based API, and a Docker image for self-hosting. The MCP server is the interesting one for agent workflows: it lets an AI agent query your databases through the Model Context Protocol, turning DBX from a human tool into an agent tool.