What to compare
| Capability | Why it matters |
|---|---|
| Request-level logs | Needed for debugging prompt/output issues and incident root cause |
| Cost + usage + latency + errors | Prevents local optimization on cost only |
| Project/workspace attribution | Creates ownership for engineering and finance |
| Alerts and anomaly detection | Stops overspend and degradation earlier |
| Multi-provider visibility | Enables routing and comparison decisions |
