The backend for apps and agents, built to scale on Lakebase Postgres.
Neon is a serverless Postgres platform that markets itself as a backend for both applications and AI agents. Alongside the database it bundles Managed Better Auth, an S3-compatible Object Storage, Neon Functions, a Data API and an AI Gateway, with git-like database branching, autoscaling and scale-to-zero as the headline database features.
Status: Live and actively developed.
Published by the vendor themselves — not community forks.
| Plan | Price | Included | Notable |
|---|---|---|---|
| Free | $0/month | 100 CU-hours monthly per project, 0.5 GB storage per project, 5 GB public egress, 1-day monitoring retention Unlimited team members | 100 projects, 10 branches per project, Compute sizes up to 2 CU (8 GB RAM), Scale to zero after 5 minutes inactive |
| Launch | Usage-based - typical spend $15/mo (based on intermittent load, 1 GB) | $0.106 per CU-hour, $0.35 per GB-month storage, $0.002 per branch-hour, 500 GB egress per project then $0.10/GB Unlimited team members | 100 projects, 10 branches per project, Compute sizes up to 16 CU (64 GB RAM), Instant restore history up to 7 days, 100 snapshots at $0.09/GB-month |
| Scale | Usage-based - typical spend $701/mo (based on high load, 100 GB) | $0.222 per CU-hour, $0.35 per GB-month storage, $0.002 per branch-hour, 500 GB egress per project then $0.10/GB, $0.01/GB private network transfer Unlimited team members | 1,000+ projects, 25 branches per project, Compute sizes up to 56 CU (224 GB RAM), Instant restore history up to 30 days, configurable scale-to-zero |
| Agent Plan / Startup Program / Open Source Program | Custom - 'Learn more' links, no published figures | Agent Plan: custom rates and limits for thousands of databases. Startup Program: up to $100k in credits. | Aimed at app-generation platforms, early-stage startups and open-source projects respectively |
Checked 2026-08-07 against their pricing page. Vendors change pricing without notice.
Neon is a managed Postgres service built on a decoupled compute/storage architecture Neon calls Lakebase. That separation is what enables its two signature features: autoscaling compute independent of stored data, and instant copy-on-write branches used for dev, test and CI. Around the database Neon has added the rest of a backend - auth, object storage, functions, an HTTP data API and a model gateway - and has explicitly aimed the whole surface at AI agents, with a one-command `npx neon init` setup, an official MCP server at mcp.neon.tech, a CLI, and a REST API the docs describe as 168 endpoints across 19 resources.
Teams building app or agent backends on Postgres who want per-environment database copies without paying for idle capacity - startups and product teams using branching for preview environments and CI, and larger engineering orgs that want SLAs, HIPAA, SOC 2 and private networking without a fixed platform fee.
Pricing is entirely usage-based above the Free plan, and Neon's own pricing page publishes example spends of ~$15/mo on Launch (intermittent load, 1 GB) and ~$701/mo on Scale (high load, 100 GB) - so a workload with steady, high, predictable load is exactly the shape where per-CU-hour billing works against you versus a fixed-size instance. The Free plan is capped at 0.5 GB storage per project, 100 CU-hours per project per month and computes up to 2 CU, and support there is Discord-only, which rules it out for anything with an incident response requirement. Four of the surrounding services - Object Storage, Functions, AI Gateway and the Data API - are still labelled Beta on Neon's own docs and pricing page, so teams that need the full bundled backend under a production SLA today are buying into beta surface.
No curated rival set covers this tool yet, so these are simply other tools in the same architectural layer and product form. Treat it as a starting point, not a shortlist.
The product is actively shipping - the changelog dated 31 July 2026 announces project-level permissions.
Yes. The Free plan is $0/month with no time limit and no credit card, giving 100 projects, 100 CU-hours per project per month, 0.5 GB storage per project, computes up to 2 CU (8 GB RAM) and unlimited team members. Support on Free is community Discord only.
Usage-based with no fixed platform fee. Compute is billed per CU-hour ($0.106 on Launch, $0.222 on Scale), storage at $0.35 per GB-month, additional branches at $0.002 per branch-hour, snapshots and backups at $0.09/GB-month, and public egress at $0.10/GB after a 500 GB per-project allowance.
Yes. The manifest supports a read-only mode via an `x-read-only` header. The individual tool names are not listed in that manifest, so we have not recorded them.
Neon's API reference states 168 endpoints across 19 resources, covering projects, branches, compute endpoints, auth, organizations, API keys, snapshots, the Data API, buckets, functions, logs and consumption. The docs say everything possible in the Neon Console is possible via the API.
Postgres, Auth, branching, restore and the API are presented as generally available. Object Storage, Functions, the AI Gateway and the Data API are all labelled Beta on both the docs index and the pricing table, where they are listed as carrying no charge during beta with usage limits.
Neon describes copy-on-write branches as editable database copies created instantly for dev, test and CI/CD, with data anonymization to mask sensitive values for safe sharing, and ephemerality so obsolete branches delete themselves. Free and Launch plans allow 10 branches per project, Scale 25.
Neon's llms.txt states it is fully compatible with any language, framework or ORM that supports Postgres, and the docs list framework guides for Next.js, Django, Drizzle, Express, NestJS, Astro, SvelteKit, Nuxt, Laravel, Rails and 20+ more. Connection pooling is built on pgBouncer with up to 10,000 connections.
The pricing table lists HIPAA compliance and SOC 2 report access as included on Scale, along with private networking, IP allow rules, SSO and log/metric export.
This page tells you whether Neon Postgres fits. Cheetah builds and runs the system it goes into — enrichment, sequencing, CRM and the joins between them.
Book a call →The directory stays free. There is nothing gated behind this.