Ncurator Product Privacy Notice

Learn how the Ncurator website, desktop app, and hosted model services handle data.

Desktop app and model services

The desktop app manages chat history locally. When you use a hosted or free model provided by Ncurator, we forward the content needed to generate a response to the provider for the selected model. Services used for free models include Zhipu AI’s BigModel Open Platform, ModelScope Inference, and the DGBMC API. Model providers handle the data they receive under their applicable terms. Our backend does not persist chat content, model responses, attachment contents, or tool output. When you use your own API key, the desktop app connects to your configured provider. Your key, custom endpoint, and manually entered model names are stored locally.

Product statistics and website analytics

The released desktop app sends a random installation identifier, platform, architecture, version, app launches, accepted messages, and main task outcomes. A failure carries a category code. A completed task may include the model source and public catalog entry. We record the target platform and architecture for website download clicks. We use these statistics to understand usage and reliability. Raw events have a 365-day retention period from server receipt and are removed on the cleanup schedule. The domestic and international services store their own statistics. Applicable public pages use Google Analytics to measure site visits.

Logs and backups

Vercel hosts the international site. Its current plan keeps function runtime logs for one hour. TiDB Cloud hosts the international database. Active deletion rules are reapplied after a database restore.

To ask about data handling or request correction or deletion of personal information, email us through the contact page.