The signals the operator acts on.
Latency, cost and error per dispatch — the operator's eyes, on the record. Instead of a chart your team reads two weeks late, the operator reads every call off the same billable receipts you are metered on, then acts within your bounds: it routes around a p95 spike in two minutes and writes a receipt for the move.
The row the operator acts on is the row you're billed on
Instead of a parallel logging pipeline that drifts from your invoice, the operator senses from the receipt ledger itself. It reads latency, cost, token counts and outcome straight off the canonical receipts — so the signal it acts on, the dashboard you watch and the invoice you pay all resolve to the same rows.
No reconciliation gap to chase: one record per dispatch, every metric traces to it, and the operator's next move points back to the exact figures it saw.
- Per-dispatch signal. The operator reads latency, cost, tokens and outcome for every call.
- Provider and agent breakdowns. It slices by vendor, model, team or agent to find what to fix.
- Receipt-sourced. Every signal comes from the ledger, so what it senses reconciles with what you spend.
The operator can't sense a number you can't check
Instead of trusting a dashboard's word, you verify it: every signal the operator reads is a field on a chained receipt, so the numbers it acts on are exactly as verifiable as the receipts behind them. Recompute one in your browser below.
Every signal is a field on a receipt
The latency_ms, cost_usd and outcome the operator reads are fields on each canonical receipt. Recompute one leaf and you've confirmed a data point it acted on — no Agentics call, no taking our word for it.
This is the same client-side verification shipped on the Trust portal. SHA-256 runs via Web Crypto. No Agentics call.
- •Recompute leaf
- •Chain integrity
- •Fold Merkle proof
- •Anchor memo == root
SHA-256 runs in your browser via Web Crypto. No Agentics call.
Read the same signal the operator reads
Pull per-provider latency and cost for a window — the exact figures the operator senses, sourced from billable receipts.
curl "https://agentics.you/api/metrics?\
group_by=provider&metric=p95_latency,cost_usd&window=24h" \
-H "Authorization: Bearer ak_…"
# → [{ provider, p95_latency_ms, cost_usd, dispatches }, ...]
# the operator acts on these same figures; each traces to a billable receiptSee the signal, then the move it triggered
The console charts latency, cost and error rate per provider and agent, refreshed from the ledger — and links each signal to the action the operator took on it.
Signal Metrics console
Per-provider latency, cost and error-rate charts, each derived from the same billable receipts and linked to the operator's response.
Open Signal Metrics →Give the operator eyes you can verify.
Observe & Suggest is available now; assisted and autonomous tiers roll out per tenant as you grant them. The operator runs live and proves every move.