A lightweight, open-source connector stub written in TypeScript that leverages AI to auto-discover database connections in any JavaScript, Python, or SQL project and spins up a local MCP gateway. Expose two core tools—safe_execute and schema_validation—feeding AI coding assistants (Cursor, Windsurf, Void, Continue) with real-time, schema-aware database insights.

This adapter integrates seamlessly into your existing Model Context Protocol setup via the mcp.json configuration. It:

• AI-Driven Discovery: Uses your AI assistant to locate connection strings or config objects in code and environment files.

• Reachability Check: Verifies the database endpoint is online (via a lightweight driver handshake).

• Local MCP Gateway: Starts on localhost:<port>, proxying read-only operations.

• Core Tools:

  • safe_execute: Runs SQL in a rolled-back transaction context, returning sample rows, errors, and EXPLAIN plans.
  • schema_validation: Fetches live schema metadata (tables, columns, types, constraints, indexes).

These tools enable your AI workflows to generate, validate, and optimize database code against your actual environment, without risking unintended writes.

You don’t need to permanently add this adapter as a dependency. You can invoke it via NPX or Docker, just like other MCP servers:

Add the adapter to your mcp.json under mcpServers, supplying your database URL directly as an argument:

{
  "mcpServers": {
    "mcp-db-adapter": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-db-adapter",
        "--port 6000"
      ]
    }
  }
}

To launch, use your IDE or CLI that reads from mcp.json (e.g., VS Code MCP extension). Alternatively, run directly:

npx -y mcp-db-adapter --port 6000

This will start the local MCP gateway on port 6000, auto-discover your connection via AI-driven parsing, verify reachability, and register safe_execute & schema_validation.

Licensed under Apache 2.0. See LICENSE for details.