I Ported Python's Best PostgreSQL Tools to TypeScript

Python has excellent PostgreSQL tooling — schemainspect, migra, pg_activity. TypeScript has... not much. So I ported them.

The Problem

If you're a Node/TS developer working with PostgreSQL, your options are:

• Schema inspection? Write raw information_schema queries or use Prisma's heavy ORM
• Schema diff? Nothing. You compare manually or switch to Python
• Health checks? Copy-paste SQL snippets from blog posts
• Activity monitoring? Open pgAdmin or SSH into the server

Python developers have had elegant solutions for years. TypeScript developers deserve the same.

What I Built

pg-toolkit — a unified CLI and SDK that brings 5 PostgreSQL tools together:

1. doctor — 20+ health checks with scoring and auto-generated fix SQL
2. inspect — Full schema inspection (tables, views, functions, indexes, enums, triggers, RLS...)
3. diff — Compare two databases, generate migration SQL (ALTER/CREATE/DROP)
4. top — Real-time activity monitor (like pg_activity but in your terminal)
5. types — Generate TypeScript types from your database schema

One npx command. No Python. No Docker. No config files.

The Interesting Part: Porting Decisions

schemainspect → pg-inspect

• Original: 2000+ lines of Python, deeply coupled to SQLAlchemy
• My version: Pure SQL queries, returns plain objects, zero ORM dependency
• Lesson: The hardest part wasn't the SQL — it was understanding which information_schema views actually give you complete information (spoiler: they don't, you need pg_catalog)

migra → pg-diff

• Original: beautiful Python code by Robert Lechte
• Challenge: migra's diff algorithm is surprisingly subtle — handling dependency order (can't drop a type before the column that uses it) took 3x longer than expected
• Safe mode: --safe flag that never generates DROP statements. Because production.

pg_activity → pg-top

• Original: ncurses-based Python TUI
• My version: Uses lipgloss-ts (itself a port of Go's lipgloss) for styling
• Meta moment: A TypeScript port using a TypeScript port of a Go library to display PostgreSQL stats

Why Not Just Use the Python Originals?

Fair question. If you're already in a Python environment, those tools are great. But:

1. Your deploy pipeline is probably Node-based — adding Python just for DB tools is friction
2. Programmatic access — import { doctor } from '@indiekitai/pg-toolkit' works in your existing codebase
3. MCP support — Every tool works as an MCP server, so AI agents (Claude, Cursor) can inspect and diagnose your database directly
4. Single dependency — One npm install instead of managing Python virtualenvs

The MCP Angle

This is what I'm most excited about. Every tool in pg-toolkit exposes an MCP server. Your AI coding agent can:

• Inspect your actual database schema before writing queries
• Run health checks and suggest optimizations
• Compare staging vs production schemas
• Monitor running queries

Not "AI-powered" (no LLMs inside). Just well-structured tools that AI agents can use.

Numbers

• 7 packages, all published on npm under @indiekitai/
• ~100 tests across the suite
• Works with PostgreSQL 12+
• Zero native dependencies (pure TypeScript + node-postgres)

Try It

npx @indiekitai/pg-toolkit doctor postgres://localhost/mydb

GitHub: https://github.com/indiekitai/pg-toolkit · npm: https://www.npmjs.com/package/@indiekitai/pg-toolkit

Feedback welcome. This is v0.2 — rough edges exist.