Recovery & Delete Control is now in early access: use it on your own database nowuse it now

Data that outlives its operators.

Every operator can fail. Your data should not go with it.

Your data depends on operators you do not fully control.

Companies now hand their data to operators they do not fully control: cloud providers, platforms, AI agents, and the vendors who build their tools. Any of them can lose that data, destroy it or be compelled to hand it over. We are building DataPrism so that a company's data survives every one of its operators, us included.

Each sits between a company and its data for a good reason: a cloud stores the data, a platform runs it, an agent changes it, a vendor builds the tools around it. Each is also a way for that data to disappear.

A suite on one common base.

DataPrism is a suite of products on one common base. Together, they are designed to let a company:

In early access

Hold a verified copy

Out of reach of the credentials that could delete it, restored and checked on a regular schedule.

Decide who may destroy

Destructive operations wait until a verified copy covers them.

Next, on the same base

  • See which credentials can destroy

    A local check that shows which tokens could delete a database or its backups.

  • Recover without the first provider

    A second copy at another provider, with a failover that is tested.

  • Erase everywhere

    Apply an erasure request to every copy, and show that it was applied.

  • Prove each step

    Signed attestations that a third party checks without a DataPrism account.

  • Spread files across providers

    Encrypted on your side, so that none of the providers holds a readable version and losing one costs no file. This is the idea DataPrism started from.

  • Give an agent access for one task

    Then take it back when the task ends.

  • Let agents operate through signed intents

    Infrastructure changes executed under your authority.

Production databases, first.

We start where the risk is sharpest today: production databases changed by AI agents, scripts and CI, so that one wrong command is never final. Those are the first two capabilities above, in early access now.

The rest of the suite is built on the same base, one capability at a time, and each promise covers only what a customer has installed.

The same rules, whatever we add.

No key, no role
Throughout the suite, DataPrism holds no key, no role in your accounts and no right to delete.
Your cloud
Everything is installed in your own cloud, with your own rights.
Provider-agnostic
Provider-agnostic by design. We start with Supabase, Neon and AWS, add providers through adapters, and state each promise provider by provider.
Evidence
Each step is meant to leave evidence that a third party can check without asking us.

Start with your production database.

Recovery & Delete Control is in early access: a verified copy of your production database, and destructive changes that wait for it. Pricing on request.