pgmt — PostgreSQL Schema-as-Code

Your schema files are the source of truth. pgmt tracks dependencies, applies changes instantly, and generates production-ready migrations.

One function changed. pgmt detected 3 dependent views, dropped them in the right order, applied the update, and recreated everything — automatically.

# Shell (macOS/Linux)
curl -fsSL https://pgmt.dev/install.sh | sh

# npm
npm install -g @pgmt/pgmt

# Cargo (from source)
cargo install pgmt

# Initialize (new project or from existing database)
pgmt init
pgmt init --dev-url postgres://localhost/my_existing_db

# Edit schema files, then apply to dev
pgmt apply

# Generate migration when ready to ship
pgmt migrate new "add user analytics"

# Deploy to production
pgmt migrate apply --target-url $PROD_DATABASE_URL

# Provision a brand-new environment (demo, staging, DR) from a baseline
pgmt migrate provision --target-url $NEW_ENV_DATABASE_URL

Edit a schema file, save — your dev database updates instantly. No migration files during development, no manual dependency management. The database finally works like the rest of your development environment.

$ pgmt apply --watch
👀 Watching schema/ for changes...

  schema/functions/calculate_score.sql changed

📋 8 changes
  ✓ Drop view public.executive_dashboard
  ...
  ✓ Create view public.executive_dashboard

✅ Applied 8 changes

When you're done iterating, pgmt migrate new generates an explicit SQL migration file. You review it, edit it if needed, and deploy it. Nothing touches production without your approval.

Multi-section migrations handle the hard stuff — concurrent indexes, data backfills, different timeouts — all in one file with per-section retries:

-- pgmt:section name="add_column" timeout="5s"
ALTER TABLE users ADD COLUMN verified BOOLEAN;

-- pgmt:section name="backfill" mode="autocommit" timeout="30m"
UPDATE users SET verified = false WHERE verified IS NULL;

-- pgmt:section name="add_index" mode="non-transactional" retry_attempts="10"
CREATE INDEX CONCURRENTLY idx_users_verified ON users(verified);

• Quick Start — Get up and running
• Adopt Existing Database — Import existing schema
• CLI Reference — All commands and options
• Blog — Why schema-as-code?
• Changelog — Notable changes in each release

• PostgreSQL 13+ (tested on 13–18)
• Rust 1.85+ for building from source

./scripts/test-setup.sh   # One-time setup
cargo test                 # Run tests

See the contributing guide for details.

MIT — see LICENSE.