Privacy
Clear boundaries, by design.
This page separates the public website from the desktop app because they handle different kinds of data.
The public website
This static website does not include account forms, payment forms, cookies, or analytics. It requests the latest public release metadata from GitHub so it can show only installers that actually exist.
| Request | Why | Destination |
|---|---|---|
| Latest release metadata | Version, publication date, installer name, size, and download URL | GitHub API |
| Installer download | Deliver the file selected by the visitor | GitHub Releases |
The desktop app
JIWWWWW runs managed services on localhost. Its Rust proxy does not persist prompt or request bodies. It reads Headroom response metadata and records activity plus daily aggregates in a local SQLite database inside the platform application-data directory.
When the optional Memory & SSOT add-on is enabled, Headroom may derive evidence-backed facts and recovery patterns from requests and store them locally. Stable facts live in ~/.codex/memory/FACTS.md and project decisions in the project's docs/DECISIONS.md. Full-message logging remains disabled, and the add-on policy forbids storing secrets or raw prompts in those files.
- Authorization headers, API keys, session tokens, and full prompts are excluded from logs and diagnostics.
- Diagnostics exports exclude database contents, request bodies, prompts, and credentials.
- Codex and Claude Code credentials remain owned by their clients.
- Local logs rotate on the device.
Accounts and billing
Signing in is optional for local optimizer features. When configured, the desktop app uses Supabase for email/password accounts and subscription status. Supabase manages password hashing; this app does not store passwords in its application tables. The Supabase SDK persists the session in the desktop webview profile.
Stripe is the source of truth for billing. Payment details are entered on Stripe-hosted pages and are not stored in the app's subscription table.
Upstream providers
Optimization does not make model requests offline. Compatible requests are forwarded from the local runtime to the upstream provider configured for each enabled client. Those providers' own privacy terms still apply.
Questions
Use the support page to report a privacy concern.
Last updated: September 15, 2026.