Gaz MCP DB

gaz-mcp SQL

Use sql_query to inspect a configured MySQL or PostgreSQL environment. The tool is intentionally read-only.

Tool contract

Call sql_query with:

The result contains columns and rows, with row values encoded as strings.

Workflow

1. Read the tool description and select an advertised environment. Never guess its name.
2. Discover the schema before assuming table or column names.
3. Run the smallest query that answers the question.
4. Interpret the result with the engine's semantics in mind.

For MySQL, start with:

sql_query(environment="development", database="app", query="SHOW TABLES")
sql_query(environment="development", database="app", query="DESCRIBE users")

For PostgreSQL, use catalogs:

sql_query(environment="analytics", database="reporting", query="SELECT tablename FROM pg_catalog.pg_tables WHERE schemaname = 'public' LIMIT 50")
sql_query(environment="analytics", database="reporting", query="SELECT column_name, data_type FROM information_schema.columns WHERE table_schema = 'public' AND table_name = 'users' ORDER BY ordinal_position")

Then query explicit columns with a limit:

sql_query(environment="development", database="app", query="SELECT id, email FROM users ORDER BY id DESC LIMIT 20")

Operating rules

• Add LIMIT unless the complete result is necessary.
• Prefer explicit columns over SELECT *.
• Use EXPLAIN before making performance claims; do not use mutating variants such as EXPLAIN ANALYZE when they could execute writes.
• Keep exploratory queries focused because environments use a small connection pool.
• Do not attempt DDL, DML, transactions, session changes, stored procedures, or administrative commands. The proxy and database session enforce read-only access.
• Never claim a table, relationship, or performance cause that the returned schema or plan does not support.