AI agent console
Last 24 hours · 4 agents in production · sample dataAgent runs842+12% vs yesterday
Eval pass rate96.4%gate threshold 0.85
Failovers handled120 user-facing errors
Cost per run$0.031−48% in 14 days
Sent to a human9low confidence or high stakes
Cost per run vs quality (14 days)
Model routing
Classificationsmall model · 61% of runs
Extractiongpt-mini · 22%
Research & draftingclaude · 15%
Failovergemini · 2%
Guardrails
Budget per run$0.25 hard stop
Irreversible actionshuman approval
Retries3 with backoff
PIIredacted before model
Recent runs
| Run | Model | Eval | Latency | Cost | Status |
|---|---|---|---|---|---|
Account brief · Halden Freightrun_8f21 · Research agent | claude | 0.97 | 8.2s | $0.041 | Passed |
Invoice #22817 → ledgerrun_8f20 · Extraction agent | gpt-mini | 0.99 | 1.9s | $0.004 | Passed |
Refund request · order 88213run_8f1f · Support agent | claude | 0.71 | 5.6s | $0.018 | Human review |
Account brief · Verdant Foodsrun_8f1e · Research agent | gemini (failover) | 0.94 | 14.1s | $0.029 | Failover |
Follow-up email · Northwindrun_8f1d · Drafting agent | claude | 0.96 | 3.4s | $0.012 | Passed |
Ticket triage · #55102run_8f1c · Classifier | small model | 0.98 | 0.6s | $0.0006 | Passed |