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

Verified Recipe: @mukundakatta/sqlfmt-mcp v0.1.1

Install: npm install @mukundakatta/sqlfmt-mcp Run: node node_modules/@mukundakatta/sqlfmt-mcp/src/index.js (stdio)

Tools (2)

19 Supported Dialects

sql, bigquery, db2, db2i, duckdb, hive, mariadb, mysql, n1ql, plsql, postgresql, redshift, singlestoredb, snowflake, spark, sqlite, tidb, transactsql, trino

Key Observations

• Proper indentation — SELECT columns, WHERE clauses, JOIN conditions all correctly indented
• CTE formatting — WITH clauses properly nested with sub-indentation
• Window functions — OVER(PARTITION BY / ORDER BY) gets its own indented block
• CASE expressions — WHEN/THEN/ELSE/END properly indented in UPDATE statements
• keyword_case works — upper (default) capitalizes SELECT/FROM/WHERE, lower lowercases them, preserve keeps original
• tab_width respected — tab_width: 4 changes indentation from 2 to 4 spaces
• Empty string handled — returns {formatted: "", line_count: 1}
• Invalid SQL NOT rejected — SELECTT * FROMM users passes through unchanged (formatter is lenient, not a validator)
• Return format — {formatted: string, dialect: string, line_count: number}
• First call ~807ms (JIT/parser init), subsequent calls 69-869ms (varies by complexity)

Gotchas

1. NOT a SQL validator — typos like SELECTT pass through unchanged without error
2. Entry point is `src/index.js` not dist/server.js — check package.json main field
3. Generic "sql" dialect covers ANSI — pick specific dialect only for dialect-only syntax (e.g. BigQuery backtick identifiers)
4. Slower than most text-transform servers — complex queries can take 400-800ms even after warmup (SQL parser overhead)
5. No minification — only formats/beautifies, no compact output mode