Everything stays local
CPU, GPU, memory, disk, and network are measured on-device and never leave your Mac.
Privacy
Local by default. Network only when you ask.
SpinCat reads system metrics with first-party macOS APIs. There is no analytics, no telemetry, and no background network for system monitoring. The only outbound traffic comes from the AI usage feature you explicitly enable.
The short version
No sudo or subprocess
No shelling out, no powermetrics, no PTY probing, no arbitrary subprocesses — ever.
No analytics
SpinCat collects no usage data, sends no telemetry, and has no tracking of any kind.
Opt-in network
AI quota requests run only when you enable AI usage and turn on a specific provider.
What each metric uses
| Area | What SpinCat uses | What it avoids |
|---|---|---|
| CPU | Mach host_processor_info tick deltas | ps, process ranking |
| GPU | IOKit IOAccelerator known utilization keys | sudo, powermetrics, full IORegistry dumps |
| Memory | host_info, host_statistics64, pressure/swap sysctl | top, per-process memory |
| Disk | Root volume capacity, IOKit storage byte counters | SMART, full mount path / UUID logging |
| Network | SystemConfiguration interface, getifaddrs counters | Public IP, ping, SSID / IP logging |
| AI usage | URLSession quota/OAuth for enabled providers, localhost callback | Provider CLIs, PTY probing, arbitrary subprocesses |
How credentials are stored
- ChatGPT/Codex, Claude, and Antigravity OAuth use SpinCat-managed local credential files.
- Copilot tokens and API keys are stored in the macOS Keychain.
- Cursor is read through a security-scoped bookmark you approve, or another allowed local source.
- The quota cache stores display-only snapshots — never tokens or secrets — and restores them as stale.
Sandboxed & transparent
SpinCat runs inside the App Sandbox. Entitlements are limited to a network client for optional AI quota and OAuth, plus a localhost server for OAuth callbacks. It ships with a privacy manifest and declares that it uses no non-exempt encryption.