Patent-pending cross-database correlation

The stone that holdsthe others in place.

An arch does not stand until the keystone is set. Your data is the same: separate databases, each sound on its own, none of them answering the question you actually have. Keystone is the piece that locks them together — reading every schema, finding the join, and returning one correlated answer, without your databases ever connecting to each other.

Read the technical briefEvery claim on this page can be checked. The brief shows you how.
The keystone carries no more weight than any other stone. It is simply the one without which nothing holds.

Any tool can query one database. The hard part is querying several at once, proving the join is right, and never letting the data leave.

I

It finds the join itself

Keystone reads every connected schema and works out which columns actually correspond — across PostgreSQL, MySQL and MongoDB. You do not write the join. You are told, as a number, how confident it is in the one it found.

Ask in plain English. The correlation is the platform’s job, not yours.

II

It cannot write

Writes, schema changes, privilege escalation and injection attempts are refused before they reach the connection. Underneath that, every statement runs inside a read-only transaction — so even if the first layer were bypassed, the database itself rejects the write.

The safety is structural. There is no code path that opens a writable transaction, so there is no setting to turn off.

III

It runs inside your walls

Deploy Keystone in your own environment and your results never cross your boundary. Residency stays where your regulator expects it. Retention stays your policy, on your schedule.

Answered by architecture, so it does not depend on trusting us.

03–06

Four more, and the way to check all six

Everything above is checkable. The brief sets each claim beside the exact thing you can do to prove it — including the two places where the honest answer isn’t the flattering one.

The technical brief

100
Rows correlated
100%
Join confidence
2
Databases joined
0
Writes possible
From a real query run on the platform, not an illustration. The run is on the record in the brief.

See it join two live databases, and ask it something of your own.

It takes about thirty minutes, on your data or ours.