SQL Database Connector FAQs

Question 1

How does the connector ensure safety when LLMs query databases?

Accepted Answer

Safety features include read-only access by default, automatic row limits (default 100), query timeouts (default 30s), and credential redaction. For PostgreSQL/MySQL, it also uses `SET TRANSACTION READ ONLY`.

Question 2

What tools are available for LLMs to manage and query databases?

Accepted Answer

LLMs can utilize structured tools for database introspection (e.g., `list_databases`, `list_tables`, `describe_table`, `sample_data`) and to execute arbitrary SQL queries via the `query` tool, returning results as JSON.

Question 3

Can LLMs perform write operations (INSERT, UPDATE, DELETE) using this tool?

Accepted Answer

By default, the connector enforces read-only access. Write operations are disabled but can be explicitly enabled by setting the `--allow-write` flag during configuration.

Question 4

What is the SQL Database Connector (mcp-sql)?

Accepted Answer

The SQL Database Connector (mcp-sql) enables Large Language Models (LLMs) to securely connect and interact with PostgreSQL, SQLite, and MySQL databases, facilitating data retrieval and management tasks.

Question 5

Which databases does the SQL Database Connector support?

Accepted Answer

It provides full support for PostgreSQL, SQLite, and MySQL databases. Connections can be established using standard URLs or environment variables.

SQL Database Connector

SQL Database Connector

主な機能

ユースケース

主な機能

ユースケース