oowoo

oowoo builds and maintains a formal model of what an organisation's data means — the customers, accounts, products and policies it deals in, how they relate, and the rules that hold between them. Databases and documents, one model.

A schema is not a model of meaning. It records that a column exists and what type it holds — not what a customer is, when an account counts as closed, or whether billing and support mean the same thing by it. That knowledge exists, in people and tickets and the logic of a thousand queries, just nowhere a machine can read it. An ontology is where it becomes explicit — in a form a machine can check and a person can argue with.

Most tools here are build-once: they scan a source, produce a model, and it starts going stale the moment a column is renamed or a document is superseded. The symptoms are familiar — two dashboards that disagree, a number nobody will sign off on, a quarter spent working out what a table means before it can be moved.

That drift used to be survivable. A person started every question, and a person noticed when the answer came back wrong. Increasingly the question starts with an agent, and nobody is reading over its shoulder.

A wrong answer is not caught. It is acted on.

So the model cannot stay a document people consult. We run it as an MCP server your agents and tools query directly — so they answer from what the business agreed a thing means, not from a guess at what a column name implies.

A model that changes continuously has to be observable while it does: what changed, when, and on whose say-so; which parts are settled and which are still inference; what an agent was working from when it answered. A semantic layer nobody can audit is a second system of record with nobody accountable for it.

Models will keep getting better at understanding your data. They will never know what you decided about it. Holding on to that distinction — what was derived, and what was decided — is continuous work, and it is what we are building.