FAQ

Isn't this just Neon?

We share Neon's lineage — separated storage and compute, copy-on-write branching — and we are wire-compatible with it, so stock Neon clients and drivers work against Kisenon unmodified. That is deliberate, and it is not going away.

Below that line we diverged. The storage engine is a hard fork on its own release path, and the control plane, scheduler, billing and operational tooling are unique implementations in Go. The clearest evidence of the divergence: Kisenon runs Postgres 18, while the open-source Neon storage layer is still on 17 — upstream's engine development moved into private repositories, so there is no newer public release left to consume. We do the major-version ports ourselves.

The part with no Neon equivalent is Agent-Safe Change Control. Every statement an agent proposes is risk-classified before it can land, and a server-side gate refuses the promote when your policy says that class may not land — a human clicking Approve cannot override a block. Neon gives you a branch. Kisenon gives you a branch an agent cannot escape.

Is it safe to let an AI agent change my database?

Yes — that's what Kisenon is built for. An agent works in a scoped, budgeted sandbox fork. The fork is a copy-on-write branch, so the agent gets a full writable database without copying your data and without touching main. Its changes are risk-classified and blast-radius-checked, promoted to main only through a server-side gate, and reversible with one command. It never holds a credential that can write production. See Agent-Safe Change Control for the whole model, and Common pitfalls before you wire an agent up.

What region is data stored in?

In the region you picked when you created the project — see Regions for the current catalog. More regions land as demand grows. Your connection host carries the region (<endpoint_id>.<region>.kisenon.com), so it's derived, not something you hardcode.

Is there a free tier?

Yes. Every account starts with a free monthly allowance per organization. An organization with no payment method on file can also be held to a fixed project-count cap, after which project creation returns free_allowance_exhausted. Usage (compute, storage, branches, egress) is metered and shown in the console, with escalating banners at 80%, 95%, and 100%. See Usage & billing.

How does scale-to-zero affect my application?

Idle endpoints suspend after 5 minutes. The next packet wakes them fast enough that ORMs do not see a timeout — Endpoints carries the current wake-latency figure. The very first wake of a freshly created project, or one after a long (24 h+) idle, can take longer while the page cache warms; allow your client one retry. See Troubleshooting.

If your workload is latency-sensitive on cold starts, you can keep an endpoint warm with a periodic SELECT 1.

Is the console available in my language?

Yes. The console and these docs ship in eight languages — English, Japanese, German, French, Spanish, Portuguese (Brazil), Korean, and Simplified Chinese. Pick yours at Settings → Language; the public site also follows your browser's language with no sign-in required. See Languages.

Can I use my existing Postgres tooling?

Yes. Kisenon speaks vanilla Postgres wire protocol. Anything that talks to Postgres works: psql, pg_dump, drivers, Drizzle, Prisma, SQLAlchemy, etc.

Where do I report bugs?

File issues at github.com/Seiraiyu/Kisenon or reply to your welcome email.

Where do I report security issues?

See /.well-known/security.txt for the disclosure contact.

Is there a CLI?

Yes — see CLI. Install via curl ... | sh (mac/Linux) or irm ... | iex (Windows PowerShell).

How do I check service status?

The public status site is kisenonstatus.com, hosted off-cloud on purpose (a separate failure domain), so it stays up even when the platform is degraded. It shows per-dependency health — Postgres connectivity and API availability. Platform incidents are also tracked in the public Kisenon issue tracker: github.com/Seiraiyu/Kisenon/issues. See Status for details.