Built as a linkable reference for teams setting their first AI budget alert. Cite the rules below; then configure Scout in TryTokka.
| Topic | Rule of thumb | Why | Tool |
|---|---|---|---|
| Daily threshold | Start near 1.5× a normal day’s estimate | Catches spikes without firing on every quiet day | Open → |
| Monthly threshold | Set near your projected monthly list-rate budget | Protects the invoice, not just one bad afternoon | Open → |
| Sync cadence (Hobby) | Scout syncs about once per day; Sync now for same-day totals | Alerts email after the next successful check — not real-time hard stops | Open → |
| Multi-provider | One Scout view across OpenAI, Anthropic, Gemini, OpenRouter… | Provider consoles do not combine invoices | Open → |
| Kill switch | Alerts ≠ hard traffic cut | Keep rate limits, prepaid care, or rotate keys for runaway jobs | Open → |
| Estimates vs invoice | OpenAI/Anthropic totals are usage×rate estimates | Can differ from the final provider invoice — reconcile in Accuracy status | Open → |
No. Scout emails after the next successful sync (daily by default, or Sync now). For overnight loops, use provider rate limits, prepaid balances, or rotate keys — do not treat an email alert as a hard kill switch.
Use the spike-risk estimator for a daily starter (~1.5× normal day) and the monthly projector for a monthly budget. Tighten after you see a week of real syncs.
No. TryTokka reads billing/usage endpoints with encrypted keys. Keep a gateway only if you still need routing or fallbacks.