Cost Reports

Minnie-Review | Daily 10:00 PM PT

Purpose

Track daily and cumulative API costs, compare against budget caps, flag anomalies.

Daily Cost Report Template

Each night at 10 PM, generate: YYYY-MM-DD-cost-report.md

# Cost Report — [Date]

**Minnie-Review** | [Time]

---

## Daily Breakdown

| Agent/Session | Model | Tokens In | Tokens Out | Cost | % of Daily Cap |
|---------------|-------|-----------|------------|------|----------------|
| Main session  | Sonnet 4.5 | X | Y | $Z | % |
| Morning briefing | Sonnet 4.5 | X | Y | $Z | % |
| Evening reset | Haiku 4.5 | X | Y | $Z | % |
| Heartbeat | Haiku 4.5 | X | Y | $Z | % |
| **TOTAL** | | | | **$X.XX** | **X%** |

**Budget status:** ✅ Within cap / ⚠️ Approaching cap / ❌ Over cap

---

## Week-to-Date

| Date | Cost | Cumulative | Notes |
|------|------|------------|-------|
| Mon | $X.XX | $X.XX | |
| Tue | $X.XX | $X.XX | |
| ... | | | |

**Weekly total:** $X.XX  
**Budget:** $35/week (Phase 2), $50/week (Phase 3+)

---

## Anomalies / Flags

- [List any unusual cost spikes, runaway loops, unexpected model usage]
- [Action items if budget is trending over]

---

## Integration Notes

Data sources:
- OpenClaw session logs (via gateway API or manual parsing)
- Anthropic console (manual check until API integration)

Budget Caps (from Master Plan)

Phase 2-3:

Daily ceiling: $5
Per-agent cap: $2/day
Weekly target: $35

Phase 4+:

Daily ceiling: $10
Per-agent cap: $4/day
Weekly target: $70

Phase 5+:

Daily ceiling: $15
Weekly target: $105
Monthly target: $450

Alert Thresholds

Yellow: 75% of daily cap ($3.75 in Phase 2)
Red: 90% of daily cap ($4.50 in Phase 2)
Critical: Over daily cap — immediate Telegram alert

Cost Tracking Method (Phase 2)

Manual (until Anthropic API integration):

Check Anthropic console daily at 10 PM
Log token usage per session
Calculate cost: Sonnet ($3/$15 per MTok), Haiku ($0.25/$1.25 per MTok)
Write to daily cost report
Commit to git

Automated (Phase 3+):