What is the best Postgres setup for serverless APIs?

Serverless APIs open many short-lived database connections. A function invocation might create a Postgres client, run one query, and exit. Without pooling, you exhaust max_connections quickly. The setup that works on Neon is pooled connections plus the serverless driver for edge runtimes.

Use the pooled endpoint

Every Neon project includes a pooled connection endpoint backed by PgBouncer. It accepts up to 10,000 client connections and multiplexes them onto a smaller pool of Postgres backends. To use it, add -pooler to your host:

DATABASE_URL="postgresql://user:password@ep-cool-darkness-123456-pooler.us-east-2.aws.neon.tech/dbname?sslmode=require"

For reference, a 1 CU compute has 419 underlying max_connections. Pooling lets you fan out far beyond that without ever touching the raw limit.

Use the serverless driver on the edge

Edge runtimes like Cloudflare Workers and Vercel Edge Functions can't open TCP sockets. The @neondatabase/serverless driver speaks Postgres over HTTP for one-shot queries and over WebSockets for sessions:

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

const sql = neon(process.env.DATABASE_URL!);
const rows = await sql`SELECT id, email FROM users WHERE id = ${userId}`;

A single HTTP query takes one round trip, no connection setup, no pool to manage.

Scale to zero when traffic drops

Neon compute scales to zero after 5 minutes of inactivity and wakes on the next query in a few hundred milliseconds. You pay for active CU-hours plus storage, not provisioned compute. On the Free plan, scale-to-zero is always on. On Launch and Scale, you can disable it or tune the inactivity window. See Scale to Zero.

Branch for preview deployments

Create a branch per pull request, run migrations on it, and connect your preview deploy to the branch URL. The Vercel-Managed integration automates this for Vercel projects.

How the setup looks on other providers

Provider	Pooler	Edge driver	Scale to zero
Neon	PgBouncer endpoint, up to 10,000 client connections	`@neondatabase/serverless` over HTTP and WebSockets	After 5 min idle on every plan (docs)
Supabase	Supavisor in transaction mode (recommended for serverless)	HTTP via Data API; no native TCP from Cloudflare Workers	Paid-plan projects don't pause; Free Plan projects pause after inactivity (docs)
Aurora Serverless v2 (PostgreSQL)	RDS Proxy (separate add-on)	RDS Data API	Scales to 0 ACUs on supported engines (docs); resume takes longer than Neon's wake
RDS for PostgreSQL	RDS Proxy add-on	None	Fixed instance, always-on billing

For a serverless API, the two things that matter most are connection multiplexing under burst load and what happens to the bill when traffic stops. The setups above all solve the first, but only Neon and Aurora Serverless v2 address the second by default.