ai & your data

last updated october 2026

cay only runs on pro and max, and only when you use it. free never sends your notes to a model.

what leaves caylo

when you ask cay something, we send the request and the parts of your notes needed to answer it (the current note, notes it links to, search results) to a language model. a scholar run also sends search queries to the research sources below. we send as little as the task needs.

providers

• openai — swift and scribe (language models)
• google gemini — scribe and scholar (language models)
• anthropic — scribe on max (language models)
• tavily — web search for scribe lookups and scholar
• openalex, semantic scholar, arxiv, pubmed, crossref, unpaywall — paper search and open-access full text
• hacker news, stack exchange, github, wikipedia — forum and reference search for scholar

retention and training

our policy is zero retention: we use provider plans and settings under which your content is not stored beyond processing the request and is not used to train their models, and we won’t send your notes to a provider that won’t agree to that. this is our intent and our contract requirement; check each provider’s own terms for their exact practices.

your control

• cay never changes your words silently: what it writes is labelled, and you keep or drop it.
• everything cay does can be undone, and is in your history.
• you can see and delete what cay remembers in settings › cay.