A painting. A watch. A piece of heritage. One goes in the front door; a scored, sourced answer comes back. This is the machine that does it — pick a workflow and watch it run.
Is it real? Worth what? What’s its story? The same machine answers all three.
The engines are where the work happens. Four do the work; a fifth — governance — keeps them honest. The tools run inside them, and each engine’s colour here is the same colour that tags its tools on the Tools page.
Pulls the real records and reads your photos and papers into facts we can check.
Builds the object’s story — who made it, who owned it, where it has been.
Finds the link to another collection that no one had spotted, and what it’s worth.
Weighs the evidence and returns a confidence score — or says the evidence runs out.
A person signs off, every step is logged, your data stays in your region and yours.
This is the target architecture — where the platform is going. Working in the demo today: verifying and appraising an object end-to-end, reading marks and documents, matching people and places, the cross-collection links, the confidence score, the evidence trail, and a curator’s sign-off — on five institutions and 20'000+ objects, not yet at production scale. The two workflows marked Next — syncing enrichments back to your own system, and a Verify API for others to query — are on the way. New tools slot in behind the same doors, so the shape doesn’t change as it fills in.
One object in. Five engines. A scored answer out.
Submit a photo and follow it through, live — every source named.
See it live