Database Queries

The Database MCP server enables secure SQL queries on various database types with read-only access.

Main Features

Database Queries

• query_database: Executes read-only SQL queries
  • Supports SELECT, WITH, SHOW, DESCRIBE, EXPLAIN
  • Automatically blocks modifying commands (INSERT, UPDATE, DELETE, etc.)
  • Asynchronous execution for optimal performance

Schema Information

• get_database_schema: Retrieves database schema information
  • Lists all tables and views
  • Shows columns with data types
  • Helpful for exploring unknown databases

Configuration

The database URL is provided via the X-Database-URL header. The server supports connection pooling for better performance with multiple requests.

Supported Databases

The database tool supports the following database types:

# PostgreSQL
postgresql://user:password@localhost:5432/mydatabase

# MySQL
mysql://user:password@localhost:3306/mydatabase

# MariaDB (uses MySQL driver)
mysql://user:password@localhost:3306/mydatabase

# Microsoft SQL Server
mssql://user:password@localhost:1433/mydatabase

Security Recommendations

For production use, we strongly recommend:

1. Create read-only database users:

   • PostgreSQL:

     CREATE USER readonly WITH PASSWORD 'password';
     GRANT CONNECT ON DATABASE databasename TO readonly;
     GRANT USAGE ON SCHEMA public TO readonly;
     GRANT SELECT ON ALL TABLES IN SCHEMA public TO readonly;

   • MySQL/MariaDB:

     CREATE USER 'readonly'@'%' IDENTIFIED BY 'password';
     GRANT SELECT ON databasename.* TO 'readonly'@'%';

   • SQL Server:

     CREATE LOGIN readonly WITH PASSWORD='password';
     USE databasename;
     CREATE USER readonly FOR LOGIN readonly;
     EXEC sp_addrolemember 'db_datareader', 'readonly';

2. Restrict network access to database servers using firewalls or VPNs to the following IP: 148.251.245.94

   • This is the IP address of the MCP server that accesses the database.

3. Use connection pooling for improved performance with multiple users

4. Set appropriate timeouts to prevent long-running queries