This is important for resilience. If a page server goes down, Neon only needs to replace the shard on that server instead of taking a hit on the whole dataset. Recovery in Neon is much faster compared to systems tied to a single storage node.

Neon uses storage sharding, a lower-level approach than traditional database sharding. Different parts of a table are stored on different pageservers, which is what lets Neon scale to multi-TB database sizes without ever fitting everything on one machine. Heikki explains:

We compiled the most-used branching workflows across our platform into this blog post. Perfect if you're just getting started with branching:

More than 500k branches are created every day on Neon. Teams and agents use them for dev and staging environments, PR previews, automated CI runs, time-travel recovery, and even versioned agent workflows

@github @prisma @DrizzleORM More about Copilot custom agents and how to install them:

@github @prisma @DrizzleORM The Neon Performance Analyzer helps you identify and fix slow Postgres queries Get before/after performance metrics, recommendations, and code-level fixes in an isolated branch

@github The Neon Migration Specialist helps you run safe Postgres migrations with zero downtime Create isolated branches → test schema changes → validate → apply to production, with support for @prisma and @DrizzleORM

Custom agents are live in @github Copilot, with two Neon-powered agents you can add to your workflow today:

Modernize database workflows and migrations with @neondatabase. ❇️ It can help you: • validate schema changes • avoid unsafe migrations • tune analytical workflows • optimize transformations and queries

@kirodotdev

Just shipped: Neon is now a Kiro power @kirodotdev developers can create, branch, test, and restore Postgres databases instantly right from their IDE

We wrote about the change in detail here:

We made an infra optimization that helps us keep scaling without disruption: We moved from a central CoreDNS pod in Kubernetes to local caches using NodeLocalDNS, cutting tail latencies by as much as 87%

We’ll keep expanding our Free Plan. If you have suggestions, tell us what you’d like to see next 👇 or find us in Discord:

This last point is key. The Neon platform is extremely good at running millions of projects with incredible efficiency - We separate compute and storage, so when your DB isn’t being used, compute scales to zero for you and for us - Most projects experience intermittent usage

We’re not “being generous” or recklessly absorbing costs. This expansion of our Free plan is possible because - Our platform keeps getting more efficient thanks to ongoing engineering optimizations - We’re now running on Databricks’ global infra, which reduces operating costs

“Spacious” means two things 1. Enough CPU/memory to actually run real projects, not tiny slices that can’t power real workloads 2. Enough projects to explore many ideas in parallel

Why are we increasing this limit so much lately? We want to make Neon the default place developers go whenever they need Postgres, just like you use GitHub whenever you start a project. So we need a spacious, capable, and reliable Free plan that meets developers where they are.

Number of projects you can deploy on Neon for free: 1̶0̶ ̶2̶0̶ ̶3̶0̶ ̶4̶0̶ ̶5̶0̶ 60

This snapshot-based promotion workflow gives you a safe, repeatable way to version both data and schema across many environments. Full guide: