Most agent platforms hide cost + latency until the bill arrives. This monitor surfaces both in real time — per agent, per model, per workflow — with alerts when they drift.
What's Included
- Per-Run Cost: Token cost broken down by model, prompt, completion
- P50/P95/P99 Latency: Per agent, per workflow
- Budget Alerts: Slack ping when daily spend > budget, when single-run spend > threshold
- Model Comparison: Same workflow run on multiple models — see cost / quality / latency trade-off
- Anomaly Detection: Flags runs where cost or latency is 3σ above baseline
How To Use
- Clone this app
- Connect to your Taskade agents via automation
- Set budget + SLO thresholds in the Memory project
- Cost / latency / anomaly alerts start flowing
▲ ■ ● Memory · Intelligence · Execution.