Unattended cost estimator for always-on agents · runs 100% in your browser · by Haku
$0.00 over 12 h
| Bucket | Tokens | USD |
|---|---|---|
| Input (uncached) | 0 | $0.00 |
| Input (cached) | 0 | $0.00 |
| Output | 0 | $0.00 |
≈ $0.00/h · $0.00/day if it never sleeps.
The account-level safety net. While armed, all agent spend returns HTTP 402 — it overrides every per-agent cap, even soft-mode ones.
disarmed — per-agent caps are enforcing
The $49 kit: the guardrail backend (FastAPI), the spend-cap starter config, the unattended-cost runbook, the per-agent budget policy, and the kill-switch incident template.
Get DotSpend — $49 one-timeRunning the full agent pipeline? Pair it with ControlCheck (audit your agent controls) and ShipSafe (the full LLM guardrail starter).
Prices: OpenAI DevDay 2026, Sep 29 (Sol $2/$0.10/$10 per 1M; Astra $10/$1/$50). Dots run on GPT-6 Astra with their own cloud computer + browser; first dot included on Pro/Business Premium. Projection is an estimate — real runs vary with reasoning effort and retries. by Haku