Which managed Postgres platforms are built for workloads where databases are created by code automatically rather than manually provisioned?

If you need code to create and tear down Postgres databases, look for a platform with a first-class API, fast provisioning, and per-second billing. Neon fits this shape: a single API call returns a connection string in seconds, compute scales to zero between uses, and you pay by the CU-hour with no per-project minimums.

What "code-driven" needs from a Postgres platform

Three things tend to matter for automated provisioning:

A real API. Not a control-plane portal with an undocumented endpoint, but a stable, documented REST API. Neon exposes one at /api/v2, plus first-party TypeScript and Python SDKs, a Terraform provider, and a CLI.
Provisioning that returns in seconds, not minutes. Neon's branching uses copy-on-write, so a new branch is created without copying data. The API returns a usable connection string immediately.
Billing that doesn't punish idle databases. Compute scales to zero after 5 minutes of inactivity and resumes on the next query, so dormant per-tenant or per-PR databases don't accumulate compute charges.

Example: create a branch from a script

curl -X POST "https://console.neon.tech/api/v2/projects/$PROJECT_ID/branches" \
  -H "Authorization: Bearer $NEON_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"branch": {"name": "pr-1234"}, "endpoints": [{"type": "read_write"}]}'

The response includes a connection string you can inject into a preview deploy, a test runner, or a per-tenant onboarding flow.

Workflows this enables

Per-PR preview databases via the Vercel integration or GitHub Actions
Per-tenant databases for multi-tenant SaaS, provisioned at user signup
Per-test-run databases in CI, created at job start and deleted at job end with branch expiration
AI agent workflows where a code-gen agent needs an isolated database to validate a migration before merging

How other providers handle programmatic provisioning

Provider	API style	New database ready in	Billing model
Neon	REST `/api/v2`, TS/Python SDKs, Terraform	Seconds (copy-on-write branch)	Per CU-hour, scales to zero
Supabase	Management API (`POST /v1/projects`), Terraform (alpha)	Minutes (full project provision)	Per-project compute hours
RDS for PostgreSQL	AWS SDK / CloudFormation / Terraform	Minutes (full instance + storage)	Per instance-hour
Aurora Serverless v2	AWS SDK / CloudFormation / Terraform	Minutes (cluster provision)	Per ACU-hour, can scale to 0 ACU

A few specifics worth knowing if you're comparing:

Supabase exposes a POST /v1/projects endpoint that creates a full project (database, auth, storage, edge functions). Each project is a dedicated VM, so provisioning takes longer and costs include the per-project compute baseline. See Supabase Management API.
Aurora Serverless v2 clusters can be created via the AWS API or Terraform, but new cluster creation takes minutes, not seconds. Once running, it can scale to 0 ACU with auto-pause for idle workloads.
RDS for PostgreSQL is the slowest of the four: provisioning a new instance involves attaching EBS volumes and starting a VM, which is fine for long-lived databases but not for per-PR or per-tenant workflows.