Why dbx Beats Navicat: 15 MB, 40+ Databases, Built‑in AI and MCP

I have long found database management tools either bloated, slow to start, or locked behind expensive licenses, especially on Windows and Linux. Discovering the dbx project on GitHub Trending revealed a 15 MB, cross‑platform client that supports 40+ databases, ships with AI integration and an MCP server, and is completely free under AGPL‑3.0.

Horizontal comparison (dbx vs DBeaver vs Navicat) shows dbx’s tiny installer (~15 MB vs 400 MB+ for DBeaver and 200 MB+ for Navicat), no runtime dependencies, cold‑start time under <2 s, idle memory around 80 MB, support for 40+ databases, built‑in AI, MCP support, and a free price model, whereas DBeaver relies on a JRE and Navicat requires a paid premium version for many features.

Why it fits in 15 MB : dbx is written with Rust + Tauri 2 + Vue 3. Rust powers the backend driver layer, and Tauri uses the native system WebView instead of bundling Chromium, eliminating the heavy JVM and plugin load of DBeaver and the legacy C++ codebase of Navicat.

Performance : Cold start to the main UI takes less than two seconds, idle memory stays near 80 MB, and scrolling through a 100 k‑row result set is smooth. The same query in DBeaver takes about three seconds to appear and scrolls with noticeable lag.

Supported databases include relational (MySQL, PostgreSQL, SQLite, SQL Server, Oracle, MariaDB, TiDB, OceanBase, CockroachDB, Access), NoSQL (Redis, MongoDB, Elasticsearch), OLAP (ClickHouse, DuckDB, Doris, StarRocks, SelectDB, Redshift), Chinese vendors (DM, KingBase, openGauss, GaussDB, HighGo), and additional JDBC extensions (H2, Snowflake, Trino, Hive, DB2, Neo4j, Cassandra, BigQuery, etc.).

Core features :

Query editor built on CodeMirror 6 with metadata‑aware autocomplete, syntax highlighting, auto‑formatting, nine themes, history, and snippet saving.

Data grid with virtual scrolling, inline edit coloring (green = new, orange = modified, red = deleted), and safe batch‑save preview.

Export options: CSV, JSON, Markdown, XLSX, INSERT statements, streamed via the Rust backend for large tables.

Schema tools: ER diagram generation, schema diff with automatic ALTER statements, execution‑plan visualization, column‑level lineage analysis, and database‑wide search.

File preview: drag‑and‑drop CSV/JSON/Parquet files to preview using an embedded DuckDB engine.

AI integration goes beyond a simple chat window. The embedded AI assistant can:

Explain selected SQL, flag performance issues, and suggest indexes.

Generate SQL from natural‑language descriptions and run safety checks.

Perform dangerous‑operation confirmations (DROP, TRUNCATE, etc.).

It supports OpenAI, Claude, DeepSeek, and local Ollama models, using the current schema as context to produce accurate statements.

MCP Server (Model Context Protocol) lets external AI coding tools (Claude Code, Cursor, Windsurf) connect directly to the databases configured in dbx. Start the server with npx @dbx-app/mcp-server and configure .mcp.json as shown in the article. An AI assistant can then query “users registered in the last 7 days” without manual data export.

CLI tool : Install globally via npm install -g @dbx-app/cli and use commands such as dbx connections list --json or dbx query local "select 1" --json for scripting and CI/CD validation.

Quick start :

macOS: brew install --cask dbx Windows (Scoop): scoop bucket add dbx https://github.com/t8y2/scoop-bucket then scoop install dbx Docker:

docker run -d --name dbx -p 4224:4224 -v dbx-data:/app/data t8y2/dbx

and access the web UI at http://localhost:4224.

In summary, dbx ranks at the top of open‑source database clients for speed, size, and feature set. Its 15 MB footprint, extensive database support, built‑in AI and MCP, and free licensing make it a compelling replacement for heavier tools.