Docs

How the engine works, in detail.

Authorization, architecture and data handling, how confidence is scored, and exactly what the engine covers — written to answer the questions that otherwise take a call.

Evaluating Read in any order
Read → How connections are authorizedThree patterns, and they differ in who does the work. This is the question every partner asks on a call, and the answer is per connector. Read → Architecture and data handlingWhat crosses which boundary, what is persisted, and what is not. The alternative to this page is a questionnaire and three weeks. Read → The confidence lineWhat the score means, where the line sits, and what happens to the fields that fall under it. Read → Coverage and limitsWhat the engine supports — by protocol, auth pattern and system — and where the boundaries sit.

Reference

The connector catalog — every system already built, by name and category, with which connect on a click and which need a credential. If your customer's system is in it, the mapping work is done.

The glossary — the vocabulary these pages use as settled: the Unified Data Model, connector drift, schema inference, MCP and A2A, each in a sentence before the detail.

Implementer documentation — registering an app per provider, embedding the widget, the API reference — comes when a partner is actually implementing. Nobody implements before they decide.