Live demo — sample data on a fictional agent stack. Prices are computed from real model rates; the numbers are illustrative.
Run it on your own stack →Demo workspace
Where the money goes — per agent.
This is the audit every whatmodel workspace gets: total spend, the recoverable amount, and a live feed of every saving opportunity we spot — sliced by the agent that drove it.
Saved this month
$288
Projected based on current optimizations
vs your bill
38%
Pre-whatmodel: $763/mo
Your plan
$399/mo
Pro plan — flat fee
Where the savings come from
Right-sized compute
Workloads running on a larger model than they need
$126/mo
1 workload
Smarter routing
Workloads that can move to a cheaper equivalent
$59/mo
1 workload
Tuning + caching
Workloads where caching or settings cut spend
$81/mo
1 workload
Batch + burst
Bursty workloads eligible for batch-API discounts
$22/mo
1 workload
Compute alignment
Routing
Configuration
Workload batching
Top agents by spend · 30 days
| code-reviewer | $172.74 |
| nightly-summarizer | $98.71 |
| support-bot | $69.10 |
| data-enricher | $48.70 |
| chat-api | $29.23 |
Recent alerts
preferences- warningSpend up 23% week-over-week, concentrated in one agent.
- criticalWe rolled back an optimization in your compute alignment workload — it missed your quality bar.
- infoCost-per-call in your routing workload drifted up 6% vs its 7-day baseline.
- 22:11Swap available — would save $0.0301840 msdetails
- 21:23Swap available — would save $0.007410 msdetails
- 20:29Cache opportunity — would save $0.18022 msdetails
- 19:35Swap available — would save $0.007380 msdetails
- 18:05Semantic cache opportunity — would save $0.01031 msdetails
- 16:35Swap available — would save $0.0301760 msdetails
- 14:35Observing290 msdetails
- 12:35Swap available — would save $0.007360 msdetails
- 09:35Cache opportunity — would save $0.18019 msdetails
- 06:35Swap flagged — below your quality bar2010 msdetails
- 03:35Swap available — would save $0.007400 msdetails
- 00:35Swap available — would save $0.0301690 msdetails
Run this on your own stack.
Connect your provider or run the scan from your terminal — we read what your agents actually run, price every call, and show you exactly what to switch. The audit is free.
$ npx whatmodel scan → your real audit in 15s