AnyDB MCP Server

The Universal Database Connector for AI Agents

anydb-mcp is a professional, zero-config Model Context Protocol (MCP) server that empowers AI agents to safely query SQL and NoSQL databases.

It acts as a universal bridge, allowing tools like Cursor, Claude, Gemini, and Zed to interact with your data effortlessly using a single unified interface.

✨ Features

• 🔌 Universal Support: Works with PostgreSQL, MySQL, SQLite, MongoDB, and Redis out of the box.
• 🚀 Zero Config: No configuration files required. Just run it via npx.
• ⚡ Stateless & Fast: Connections are created on-demand and closed immediately.
• 🛡️ Secure: Designed for read-only access (enforceable via DB users).
• 🧠 Smart Parsing: Automatically detects database types from standard connection URIs.

📦 Quick Start

You can run this server directly using npx without installing anything manually.

🤖 Choose Your Client

1. Open Cursor Settings > Features > MCP.
2. Click + Add New MCP Server.
3. Enter the following details:
   • Name: anydb
   • Type: command
   • Command: npx
   • Args: -y anydb-mcp

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "anydb": {
      "command": "npx",
      "args": ["-y", "anydb-mcp"]
    }
  }
}

Edit your settings.json (Cmd/Ctrl + ,):

{
  "context_servers": {
    "anydb": {
      "command": {
        "path": "npx",
        "args": ["-y", "anydb-mcp"]
      }
    }
  }
}

Add to your mcp_servers.json:

{
  "mcpServers": {
    "anydb": {
      "command": "npx",
      "args": ["-y", "anydb-mcp"]
    }
  }
}

Note: On Windows, explicitly use npx.cmd.

1. Open the MCP Servers tab (usually in the sidebar or settings).
2. Select Configure MCP Servers.
3. Add the configuration:

"anydb": {
  "command": "npx",
  "args": ["-y", "anydb-mcp"]
}

🔗 Connection URIs

The AI agent will provide these URIs automatically, or you can provide them in the chat context.

| Database | Protocol | Example URI | |----------|----------|-------------| | PostgreSQL | postgres:// | postgres://user:pass@localhost:5432/mydb | | MySQL | mysql:// | mysql://user:pass@localhost:3306/mydb | | SQLite | sqlite:// | sqlite:///Users/alex/data.db (Absolute path) | | MongoDB | mongodb:// | mongodb://user:pass@localhost:27017 | | Redis | redis:// | redis://:pass@localhost:6379 |

🛠️ Tool Usage

The server exposes one powerful tool: db_query.

Arguments:

• uri (string, required): The connection string.
• query (string, required): The SQL query, MongoDB filter (JSON), or Redis command.
• collection (string, optional): Required only for MongoDB.

Examples:

SQL (Postgres/MySQL/SQLite):

SELECT id, email FROM users WHERE created_at > '2024-01-01' LIMIT 5;

MongoDB:

{ "status": "active", "age": { "$gt": 21 } }

Redis:

GET session:12345

🔒 Security Best Practices

1. Read-Only User: ALWAYS create a dedicated database user with SELECT-only permissions (or equivalent) for the AI.
   • Postgres: GRANT SELECT ON ALL TABLES IN SCHEMA public TO ai_user;
   • MySQL: GRANT SELECT ON mydb.* TO 'ai_user'@'%';
2. Network: Ensure the database host is accessible from the machine running the MCP server.
3. Data Privacy: Be cautious. The AI has access to whatever data the query returns.

👨‍💻 Contributing

We welcome contributions! Please see CONTRIBUTING.md for details.

1. Fork the repo.
2. npm install
3. npm test
4. Submit a Pull Request.

License

MIT © Alexeev Alexandr