MCP Database Manager

You can manage CRUD operations across multiple databases with the MCP Database Manager. It centralizes data access, supports multiple database drivers, and includes permission controls so you can maintain safe, auditable data interactions from a single MCP endpoint.

How to use

You interact with the MCP Database Manager through an MCP client. Start the local server or a published MCP runtime, then configure your MCP client to connect to it. Once connected, you can perform create, read, update, and delete operations across supported databases, manage who can access which connections, and view connection lists with sensitive parts like passwords redacted in outputs.

How to install

Prerequisites: ensure you have Python installed (Python 3.8+ is typically supported by the runtime used here). You will also need an environment where you can run MCP servers with the provided runtime shims.

Option 1 — Local Development (recommended for debugging): use the development runtime to run the MCP server from your local copy. This keeps changes immediately testable.

# Start the MCP server from your local project path
# Replace the path with the actual directory where you have the mcp-database-manager source
uv run --extra all mcp-database-manager

Option 2 — Using the published package from PyPI (stable version): run the server using the published MCP runtime from the package index.

# Start the MCP server from the PyPI package
uvx --from mcp-database-manager[all] mcp-database-manager

Configuration and operation notes

Configuration is stored in a YAML file at your home directory under ~/.mcp-database-manager/config.yaml. Define connections with a name, URL, and whether the connection is read-only. Example entry:

connections:
  - name: "main_db"
    url: "sqlite:///./main.db"
    readonly: true

The server supports multiple databases via installation extras. Use specific extras to enable drivers: postgres, mysql, mssql, or all. This lets you connect to PostgreSQL, MySQL, SQL Server, or all of them.

Security and troubleshooting

By default, permission management is enabled, and sensitive information like passwords is redacted from lists of connections. If you encounter module or runtime issues, ensure you are using the provided runtime (uv) and its environment integrates with your virtual environment automatically.

Additional notes

The MCP Database Manager is designed to work with multiple databases and provide a unified interface for CRUD actions. You can extend or customize its behavior through the configured connections and by adjusting permissions per connection.

Available tools

list_connections

Returns the configured database connections with sensitive fields redacted, enabling you to see available data sources without exposing passwords.

crud_operations

Perform Create, Read, Update, and Delete actions across the enabled database connections through a unified MCP interface.

permissions

Manage access controls per connection, enabling read-only or full-CRUD permissions for different users or agents.

sqlalchemy_integration

Leverage SQLAlchemy integration to work with supported databases through a consistent API within MCP.