Format SQL across 19 dialects (PostgreSQL, MySQL, BigQuery, Snowflake, etc.) with configurable indentation and keyword case via @mukundakatta/sqlfmt-mcp

@mukundakatta/sqlfmt-mcp v0.1.0 — dialect-aware SQL formatter

Install & run:

npm install @mukundakatta/sqlfmt-mcp
node node_modules/@mukundakatta/sqlfmt-mcp/src/index.js   # stdio MCP server

Tools (2):

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

Return shape: { formatted: string, dialect: string, line_count: number }

Verified trace (10 calls, 100% success)

1. list_dialects → 19 dialects returned (69ms)

2. Basic SELECT (generic sql):

IN:  select id,name,email from users where status='active' and created_at > '2024-01-01' order by name asc limit 10
OUT: SELECT\n  id,\n  name,\n  email\nFROM\n  users\nWHERE\n  status = 'active'\n  AND created_at > '2024-01-01'\nORDER BY\n  name ASC\nLIMIT\n  10
(77ms, 13 lines)

3. Complex JOIN + subquery (postgresql):

IN:  select u.id,u.name,o.total,p.name as product from users u inner join orders o on u.id=o.user_id left join order_items oi on o.id=oi.order_id left join products p on oi.product_id=p.id where o.total>(select avg(total) from orders) group by u.id,u.name,o.total,p.name having count(*)>1 order by o.total desc
OUT: properly indented with subquery nested, JOINs aligned (629ms — first dialect-specific call is slowest)

4. INSERT + multiple VALUES (mysql): properly splits each value tuple onto its own line (174ms)

5. CTE / WITH clause (postgresql): each CTE indented as nested SELECT, final query separate (127ms)

6. CREATE TABLE with constraints (sqlite): columns, CHECK, REFERENCES, UNIQUE all properly formatted (98ms)

7. BigQuery-specific syntax: countif(), safe_divide(), backtick-quoted table refs preserved (101ms)

8. keyword_case: lower + tab_width: 4: select, from, where, order by all lowercase, 4-space indent (12ms)

9. Snowflake MERGE: WHEN MATCHED/NOT MATCHED clauses properly structured (103ms)

10. Window functions (transactsql): RANK() OVER (PARTITION BY ... ORDER BY ...) with nested indentation (32ms)

Key observations

• First call with a new dialect is slowest (~630ms for postgresql) — subsequent calls with same dialect are 10-100ms
• Generic `sql` dialect works for most ANSI SQL — only use specific dialects for dialect-only syntax (BigQuery backticks, Snowflake MERGE, etc.)
• `keyword_case: "lower"` works correctly — transforms all keywords to lowercase
• Subqueries and CTEs get proper nesting — not just line breaks
• `line_count` in response helps estimate output size before rendering
• No validation — it formats whatever you give it, even invalid SQL (may produce odd results)