cd /news/ai-agents/neon-for-agent-platforms-a-companion… · home topics ai-agents article
[ARTICLE · art-13474] src=neon.com pub= topic=ai-agents verified=true sentiment=↑ positive

Neon for Agent Platforms: A Companion Skill for the AI Agent Program

Neon released a companion skill for AI agent platforms that teaches agents how to provision and manage fleets of databases at scale. The skill covers dual-org fleet models, per-tenant provisioning, project transfer, compound checkpoints, and consumption tracking — patterns used by major customers like Replit and Retool. It addresses the gap between standard database usage and the operational needs of agent platforms that must create and manage thousands of databases programmatically.

read3 min publishedMay 22, 2026

We've been working on neon-for-agent-platforms, a companion skill that complements our existing neon-postgres skill. The existing skill covers auth, drivers, branching, the Data API, MCP, and core Postgres-on-Neon guidance. Agent platforms have a different problem though: they don't just use one database, they provision and operate fleets of them.

This companion skill fills that gap with dual-org fleets, project transfer, per-tenant provisioning, compound checkpoints, and the consumption API. It teaches agents how to create and manage thousands of databases as a platform, using real-life patterns adopted by some of our biggest agent customers like Replit and Retool.

What it includes #

The repo includes a companion Agent Skill and sample TypeScript scripts you can run directly. It covers the control-plane patterns agent platforms actually need:

Dual-org fleet model: Two Neon organizations (sponsored free-tier vs. paid), with your control plane choosingNEON_ORG_ID

per customer tier when creating projects. Org-scoped API keys per org, personal API key for cross-org transfer.Per-tenant provisioning and transfer: Programmatic project creation, one project per customer app, and project transfer into the paid org on upgrade.Dev, staging, and preview vs. production: How to model non-prod alongside production inside each tenant project — usually a production branch (root/main) plus development branches for isolated dev, staging, or previews. Branches are copy-on-write and get their own compute; patterns like shorter suspend timeouts on dev branches help keep fleet costs predictable. The skill also ties this to branch limits*(free vs paid orgs)*so you can reason about guardrails at scale.Versioning and compound checkpoints: Branch-based snapshots for workflow checkpoints, plus compound checkpoint patterns for app-level metadata beyond Neon IDs.Consumption API: Track usage across orgs and projects.Runnable samples:npm run neon:list-projects

,npm run branch -- list ,npm run consumption

,npm run auth-users -- meta ,npm run versioning-flow

. The full catalog and env var mapping can be found inMANAGEMENT_API_SAMPLES.md.

For drivers, connection strings, and everyday Neon app integration, it refers to neon-postgres and the Neon docs so it stays up to date.

How to use it #

Install both skills in your editor of choice:

Once both skills are installed, your AI assistant has the context it needs for platform-level work. Ask it to set up a dual-org fleet, provision a project for a new user, create a compound checkpoint, or pull consumption data across an org. It'll know the right API calls and patterns without you having to paste in docs.

Some example prompts:

  • I want each tenant to have prod on the main branch and a throwaway staging database. Show me how to create a dev branch from prod, run migrations against it, and tear it down or reset it without changing prod's connection string.
  • Create a new Neon project for a user who just signed up, and show me how to transfer it to the paid org when they upgrade.
  • I need to snapshot the current state of a user's project before running a migration. Create a branch checkpoint and store the metadata somewhere I can look it up later.
  • How do I set up guardrails so one tenant can't spin up unlimited branches and run up costs?
  • Pull consumption data across my org and break it down by project so I can attribute costs per customer.
  • Walk me through the full provisioning flow: user signs up, gets a database, works in free tier, upgrades to paid. What API calls happen at each step?

If you want to try the scripts outside of an editor, clone the repo and run them directly: We built this because our largest agent platform customers kept solving the same control-plane problems independently. Now that it's open, we'd love your input. Try the skill, run the scripts, and open an issue or PR if something's missing from your workflow.

Get started: neon-for-agent-platforms repo · neon-postgres skill · Neon docs. If you're building an agent platform on Neon and want to talk architecture, reach out. We're happy to pair on it.

── more in #ai-agents 4 stories · sorted by recency
── more on @neon 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/neon-for-agent-platf…] indexed:0 read:3min 2026-05-22 ·