GitHub radar
DBX: 20MB DBC Client with AI + MCP
A 20 MB cross-platform database client for 90+ databases (MySQL, Postgres, Redis, MongoDB, etc.) with an integrated AI SQL helper and an MCP server to let Claude Code + Cursor ask questions of your database.
DBX is a cross platform database client written in Rust. It’s only ~20MB (No Java, No Python, No chromium). Supports 90+ databases including MySQL, Postgres, SQLite, Redis, MongoDB, DuckDB, ClickHouse, Elasticsearch, MS SQL Server, Oracle, etc. With the built-in AI SQL helper, you can describe your needs in natural language and get the corresponding SQL statements, explanations, optimizations and verifications (e.g., security checks) of the generated SQLs. Also, it supports Claude, OpenAI, local AI models via Ollama. You can run an independent MCP server (npx @dbx-app/mcp-server) so that other MCP clients such as Claude Code + Cursor + Windsurf + … can connect to your DBX connections and access your databases with different permissions (e.g., read-only, safe write, full access). Available on macOS, Windows and Linux desktop apps. Also available as a Docker image for sharing. And there’s a web version too! Launched late April 2026. 16,000+ stars in 4 months.
Why a vibe-coder should care
With DBX, you can avoid unnecessary copy-pasting during AI code generation by granting Claude Code permission to access your database schema after deploying the MCP server and allowing Claude Code to query your database via the chat interface. With the help of the AI SQL helper, it’s easy to instruct Claude Code about the type of data you need without explicitly writing SQL statements.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Install DBX from https://github.com/t8y2/dbx — download the version for my OS per the README, configure the MCP server with `npx @dbx-app/mcp-server` and add it to the Claude Code config, then show me the list of available connections. Ask me for database credentials if needed.
Runs on a regular laptop — 20 MB installer, no Java or GPU needed.
Open on GitHub▌ More finds