What sign-in data is read
To read quotas, the app reads — only when needed — the current user's Codex sign-in file, specific Claude desktop sign-in cookies, the Claude Code sign-in (when the desktop app is absent), and the Antigravity Keychain item or the sign-in files of instances you enable. Only authentication- and organization-related records in Claude's own cookie database are read; browser history and other sites' cookies are never touched.
How credentials are used
Credentials are only used in memory and sent over HTTPS to the service they belong to; cross-origin redirects never carry them. Antigravity quota requests are limited to known Google services — an unrecognized endpoint causes an error rather than receiving credentials.
Tokens are never stored in project files, logs or preferences, and the original apps' sign-in files are never modified.
What is stored locally
Local preferences include app and instance choices, manually selected file paths, aliases you set, the Google account identifier used to detect account changes, and update settings such as automatic checks and the last check time. Account names and avatars from Google are cached in memory only and cleared on quit; avatar requests never carry OAuth credentials.
Software updates
By default the app checks the update feed over HTTPS once a day (served by aethernative.com on Cloudflare; earlier versions use GitHub) and downloads signed installers from GitHub when you choose to update. These services may receive ordinary connection information such as IP address and request headers — never quotas or credentials — and Sparkle system profiling is disabled. Automatic checks can be turned off in the menu.
Contact
For privacy questions, use GitHub Issues. Never upload credentials, cookies or complete authentication files.