💰

Fleet API Costs

April 1–16, 2026 — YoniClaw Fleet

$24,020
Total Spend
$1,501
Daily Average
27
Active Agents
21
OCI Servers

🚨 Critical Events

3 Detected

🔥 OC-Tuval Burst

Apr 5–6: Sudden spike to $912/day (2 days), then crash to $66/day. 400% above baseline — likely runaway process or experimental workload.

⚡ V2→V3 Migration Overlap

Apr 6–7: YoniClawV2 + V3 running simultaneously. Combined $7,824 — dual operation costs during migration window.

📈 Fleet Expansion Wave

Apr 10–16: AppsClaw, SecurityClaw, QuantClawV2 activated. Contributing $200+/day increase to baseline.

👑 Top Spending Agents

65% of Total
🦞 YoniClawV3
oc-yoniclaw-prod
$4,021
$251/day avg16.7% of fleet
🦞 YoniClawV2
oc-yoniclaw-prod
$3,803
$238/day avg⚠️ Should be shut down
🔬 OC-Tuval
oc-yoniclaw-prod
$2,162
$135/day avg⚠️ Burst pattern detected
📈 TradingClaw
oc-trading-prod
$2,074
$130/day avg8.6% of fleet
💚 GoodClaw
oc-goodclaw-prod
$1,952
$122/day avg8.1% of fleet
🌐 WebClaw
oc-webclaw-prod
$1,639
$102/day avg6.8% of fleet

⚠️ Cost Spike Analysis

Anomalies
📅 April 5–6: OC-Tuval Burst
Sudden spike to $912/day for 2 consecutive days, then crash to $66/day. Likely experimental workload or runaway process. Needs audit trail investigation.
📅 April 6–7: V2→V3 Migration
Parallel operation during migration. YoniClawV2 + V3 running simultaneously causing $500+ daily spike above baseline. Migration now complete.
📅 April 10–16: Fleet Expansion
New agents activated: AppsClaw, SecurityClaw, QuantClawV2. Contributing $200+ daily increase. Expected growth as fleet matures.

🖥️ Server Utilization

21 OCI Instances
🦞
oc-yoniclaw-prod
YoniClawV3, YoniClawV2, OC-Tuval
$9,986 · 41% of fleet
📈
oc-trading-prod
TradingClaw
$2,074
💚
oc-goodclaw-prod
GoodClaw
$1,952
🌐
oc-webclaw-prod
WebClaw
$1,639
🧪
oc-qaclaw-prod
QAClaw
~$1,200
📦
Other 16 Servers
AppsClaw, SecurityClaw, QuantClaw, etc.
$5,169 · combined

💡 Optimization Recommendations

6 Actions
IMMEDIATE — Save ~$200/day
Shut Down YoniClawV2
Migration to V3 is complete. No need for parallel operation. Eliminating V2 saves ~$238/day = $7,140/month.
IMMEDIATE
Audit OC-Tuval Burst Pattern
Investigate the Apr 5–6 burst ($912/day → $66/day). Implement monitoring alerts for spikes >$500/day per agent.
MEDIUM-TERM
Server Load Balancing
oc-yoniclaw-prod handles 41% of fleet costs. Distribute workloads across underutilized servers to reduce single-point risk.
MEDIUM-TERM
Model Optimization
Top 6 agents use premium models for all tasks. Evaluate Claude Haiku or GPT-4o-mini for routine/repetitive tasks. Potential 20–30% savings.
LONG-TERM
Response Caching
Implement semantic caching for repeated queries across agents. Could reduce redundant API calls by 15–25%.
LONG-TERM
Real-Time Cost Alerts
Deploy per-agent daily cost thresholds ($100/day default). Alert to WhatsApp group when exceeded. Catch runaway processes early.

📊 Monthly Projections

44% Savings Potential
$45,038
Current Trajectory
No changes
$36,000
With V2 Shutdown
−$9,038/month
$25,200
Full Optimization
−$19,838/month (44%)
$302K
Optimized Annual
vs $540K current trajectory