Clear data & agent-access policy

A privacy page that names every subprocessor, what it receives, and where it runs, rather than a generic template.

The one line that matters most: your content is never used to train AI models. The paid tiers we run on carry that commitment contractually, and the page says which provider sits behind each step so the claim can be checked rather than believed.

Content processed through the API, the CLI or MCP follows exactly the same retention and no training rules as the web app. An agent surface with looser terms would make the whole policy decorative.