Berth

A local MCP server that gives an AI assistant structured access to PostgreSQL, SQLite and MySQL databases for querying, schema inspection, migrations and backup

What you can do

Connect to a database and run SELECT queries, with an automatic row limit, or execute INSERT, UPDATE and DELETE statements subject to the active safety mode, with a confirmation token for destructive operations in admin mode. Inspect schemas: list tables, views and indexes, describe columns, show foreign key relationships, report database and table sizes, and view currently running queries on PostgreSQL. Analyze queries with EXPLAIN ANALYZE. Generate dialect-aware migration SQL by comparing a live database against target DDL or against a second live database. Create database backups and restore from them.

When to use it

Useful when an assistant needs to explore or query a relational database, understand its schema, diagnose slow queries, draft migrations between schemas, or take and restore backups, while keeping writes and destructive operations gated behind explicit modes.

Requirements

Runs as a local process over stdio, launched by an MCP client. Install the berth-mcp Python package from PyPI; MySQL support needs the optional mysql extra. Database connection strings (DSNs) are supplied through the db_connect tool. Backup and restore paths are confined to BERTH_BACKUP_DIR, which defaults to the current working directory. Backup and restore rely on the pg_dump/psql, sqlite3 and mysqldump/mysql command-line tools.