Neon is a serverless PostgreSQL platform with instant branching, autoscaling, and a scale-to-zero compute model. Because each Neon branch is an independent PostgreSQL database, UUID primary keys are the natural choice for IDs — auto-increment integers would conflict across branches. UUID v7 improves on UUID v4 by adding monotonic ordering that keeps B-tree indexes efficient across the serverless write pattern. See UUID v7 Performance Benchmarks for the index fragmentation numbers.

Enable pg_uuidv7 on Neon

-- Run in any Neon branch console or via migration
CREATE EXTENSION IF NOT EXISTS pg_uuidv7;

-- Verify
SELECT uuidv7(); -- 018fbe3a-4c5d-7b12-8f3e-1a2b3c4d5e6f

The extension is available on all Neon projects on PostgreSQL 16 and 17. No plan upgrade required.

Schema: UUID v7 default primary key

CREATE TABLE orders (
  id         uuid        PRIMARY KEY DEFAULT uuidv7(),
  customer   text        NOT NULL,
  total      numeric     NOT NULL,
  created_at timestamptz NOT NULL DEFAULT now()
);

-- UUID v7 embeds the timestamp — created_at is redundant for ordering
-- but useful for human-readable display
CREATE INDEX idx_orders_created ON orders (id);  -- B-tree on UUID v7 = time order

Because UUID v7 encodes creation time, ORDER BY id and ORDER BY created_at produce the same result. The single B-tree index on id handles both range queries and pagination. See UUID v7 for Cursor Pagination.

Neon branching workflow with UUID v7

# Create a feature branch
neon branches create --name feature/new-orders

# Branch inherits all UUID v7 values from main
# New rows in the branch get new UUID v7 values (later timestamps)
# When branch is merged/promoted, IDs never collide with main
-- In the feature branch: insert test data
INSERT INTO orders (customer, total) VALUES ('test-user', 99.99);
-- id = 018fbe3a-... (timestamp = branch creation time + offset)

-- Back on main: production data
-- id = 018fbe39-... (earlier timestamp)
-- No collision — UUID v7 monotonicity across branches

Prisma + Neon + UUID v7

// schema.prisma
generator client {
  provider = "prisma-client-js"
}

datasource db {
  provider = "postgresql"
  url      = env("DATABASE_URL")  // Neon connection string (pooled)
}

model Order {
  id         String   @id @default(dbgenerated("uuidv7()")) @db.Uuid
  customer   String
  total      Decimal
  createdAt  DateTime @default(now())
}
-- Migration: 0001_enable_uuidv7.sql
CREATE EXTENSION IF NOT EXISTS pg_uuidv7;

See UUID in Prisma for the full migration and typed ID pattern.

Drizzle + Neon + UUID v7

// schema.ts
import { pgTable, uuid, text, numeric } from 'drizzle-orm/pg-core';
import { sql } from 'drizzle-orm';

export const orders = pgTable('orders', {
  id:       uuid('id').primaryKey().default(sql`uuidv7()`),
  customer: text('customer').notNull(),
  total:    numeric('total').notNull(),
});
// db.ts — Neon serverless driver
import { neon } from '@neondatabase/serverless';
import { drizzle } from 'drizzle-orm/neon-http';
import * as schema from './schema';

const sql = neon(process.env.DATABASE_URL!);
export const db = drizzle(sql, { schema });

See UUID in Drizzle ORM for the migration setup.

Neon HTTP API with UUID v7

Neon’s serverless driver makes SQL over HTTP easy — UUID v7 cursor pagination works naturally:

import { neon } from '@neondatabase/serverless';

const sql = neon(process.env.DATABASE_URL!);

// Cursor pagination — UUID v7 as cursor
async function getOrders(cursor?: string, limit = 20) {
  if (cursor) {
    return sql`
      SELECT id, customer, total
      FROM orders
      WHERE id > ${cursor}
      ORDER BY id
      LIMIT ${limit}
    `;
  }
  return sql`
    SELECT id, customer, total
    FROM orders
    ORDER BY id
    LIMIT ${limit}
  `;
}

UUID v7 cursors are stable across Neon scale-to-zero restarts because the cursor is a value in the database, not a connection-level sequence. See UUID v7 for Cursor Pagination.

Further reading

External references

Frequently asked questions

Does Neon support UUID v7 natively?

Neon runs PostgreSQL 17, which supports the pg_uuidv7 extension via CREATE EXTENSION IF NOT EXISTS pg_uuidv7. PostgreSQL 18's built-in uuidv7() will be available once Neon upgrades to PG 18. Until then, pg_uuidv7 is the recommended path. See PostgreSQL 18 Native UUID v7.

How do I use UUID v7 with Neon and Prisma?

In your Prisma schema, set default(dbgenerated("uuidv7()")) on your id field and enable the pg_uuidv7 extension via a migration. Neon's connection pooler (PgBouncer) is fully compatible with Prisma's @prisma/client. See UUID in Prisma.

Can I use UUID v7 across Neon database branches?

Yes. Neon branches are point-in-time copies of the database — UUID v7 values generated in a branch are monotonically ordered within that branch. Since UUID v7 embeds a Unix millisecond timestamp, values from different branches will sort correctly by creation time when merged or compared, as long as the generating application clocks are synchronised (standard NTP).