Database¶

Any SQL database. Three functions. LLM-friendly output.

Database introspection and query execution via SQLAlchemy. Supports any SQLAlchemy-compatible database (PostgreSQL, MySQL, SQLite, Oracle, MS SQL Server, etc.).

Highlights¶

Connection pooling with automatic health checks
Vertical result formatting optimized for LLM consumption
Parameterized queries for safe SQL execution
Large results truncated at 4000 characters
Per-URL connection pools with 1-hour recycling

Functions¶

Function	Description
`db.tables(db_url, ...)`	List table names in the database
`db.schema(table_names, db_url)`	Get schema definitions for tables
`db.query(sql, db_url, ...)`	Execute SQL and return formatted results

Key Parameters¶

Parameter	Type	Description
`db_url`	str	SQLAlchemy connection string (required)
`filter`	str	Substring to filter table names (tables only)
`table_names`	list[str]	Tables to inspect (schema only)
`params`	dict	Query parameters for safe substitution (query only)

Examples¶

# Get database URL from project config
db_url = proj.attr("myproject", "db_url")

# List all tables
db.tables(db_url=db_url)

# Filter tables
db.tables(db_url=db_url, filter="user")

# Get schema for tables
db.schema(["users", "orders"], db_url=db_url)

# Execute queries (parameterized for safety)
db.query("SELECT * FROM users LIMIT 5", db_url=db_url)
db.query(
    "SELECT * FROM users WHERE status = :status",
    db_url=db_url,
    params={"status": "active"}
)

Source¶

SQLAlchemy Documentation

Based on¶

mcp-alchemy by Rui Machado (MPL-2.0)