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.
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.
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.
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.
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.
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
See it join two live databases,
and ask it something of your own.
It takes about thirty minutes, on your data or ours.