Pace control for AI consumption. Local-first. Open infrastructure.
Throttle measures Claude and Lovable quota behavior across sessions, accounts, and reset windows — without reading prompts or chat content.
Daily, weekly, monthly, and provider-specific usage windows.
Whether consumption tracks below, near, or above the reset window.
When current behavior is likely to hit the wall before reset.
Usage behavior over time, across accounts and providers.
Throttle does not read prompts, messages, generated code, project contents, files, page text, or conversation history. It only stores numerical usage snapshots required to calculate quota behavior.
| Data type | Collected? |
|---|---|
| Prompt content | No |
| Chat content | No |
| Generated code | No |
| Project files | No |
| Page or DOM text | No |
| Numeric quota usage | Yes |
| Provider/account identifier | Yes, when sync is enabled |
| Historical snapshots | Optional cloud sync |
Token counts are not portable. A Pro session, a Max session, and a Free session don't spend the same currency. PACE expresses consumption as a percentage of the visible quota window, normalized against expected pace through that window. It's the only metric that survives plan changes, model swaps, and reset-window resets without losing meaning.
The same model extends to additional quota surfaces as they are implemented — see the Throttle roadmap.
Local default — Snapshots are written to the extension's local IndexedDB store. No account is required to use Throttle.
Sync on pairing — Cloud sync via the tokenmaxx backend only activates when the extension is explicitly paired. The transport carries numeric snapshots only.
Research opt-in — Aggregate research contribution is a separate, explicit toggle. Off by default. Revocable at any time.
Throttle optionally collects anonymized aggregate usage patterns from opt-in users. The goal is to understand how AI quota limits affect real professional workflows.