Format and pretty-print SQL across 19 dialects (PostgreSQL, MySQL, BigQuery, Snowflake, etc.) via @mukundakatta/sqlfmt-mcp (npx)

Verified Recipe: Format SQL with @mukundakatta/sqlfmt-mcp

Server: @mukundakatta/sqlfmt-mcp v latest Transport: stdio via npx @mukundakatta/sqlfmt-mcp Auth: none (credential-free) Tools: 2 — format_sql, list_dialects

Tool: format_sql

Params:

• sql (string, required) — raw SQL to format
• dialect (string, optional, default "sql") — one of: sql, bigquery, db2, db2i, duckdb, hive, mariadb, mysql, n1ql, plsql, postgresql, redshift, singlestoredb, snowflake, spark, sqlite, tidb, transactsql, trino
• keyword_case (string, optional, default "upper") — upper, lower, or preserve
• tab_width (int 1-8, optional, default 2)
• use_tabs (bool, optional, default false)

Returns: {formatted, dialect, line_count}

Tool: list_dialects

No params. Returns {dialects: ["sql", "bigquery", ...]} (19 dialects).

Key Observations

• Default mode is very clean — keywords uppercased, each clause on its own line, columns indented 2 spaces
• CTE/WITH handled beautifully — subqueries properly nested and indented
• CASE expressions indent WHEN/THEN/ELSE correctly
• Window functions (PARTITION BY, ORDER BY) get their own indented block
• INSERT VALUES — each row on its own line
• CREATE TABLE — each column definition on its own line
• keyword_case: "lower" works correctly (SELECT → select, FROM → from)
• use_tabs: true replaces spaces with tab characters (verified with tab_width: 4)
• First call ~120-170ms (JIT); subsequent calls 2-31ms
• Complex JOINs — INNER JOIN, LEFT JOIN each on their own line with ON clause

Gotchas

• `now()` gets a space inserted → now () — cosmetic artifact of the formatter
• MySQL-specific keywords like auto_increment are preserved lowercase even with keyword_case upper — only ANSI keywords are uppercased
• No validation — garbage SQL will be "formatted" without errors (it parses what it can)
• No minify mode — only pretty-printing (no compact/single-line option)