Supernova | Dashboard

SummarySpend, savings, accuracy, and core charts CostSpend reporting and savings opportunities AccuracyOutcome coverage and improvement signals

Cost reporting

Spend and savings overview

$11.3/mo projected savings

Projected spend$17.6/mo

996 projected sessions

Total tokens3.8M

3.8K avg per session

Available savings$11.3/mo

Could cut spend by ~64%

Workflows affected5

25 cost opportunities

All workflows

Token and spend trend

19 hour buckets

Dec 4 4AM

Dec 4 5AM

Dec 4 6AM

Dec 4 7AM

Dec 4 8AM

Dec 4 9AM

Dec 4 10AM

Dec 4 11AM

Dec 4 12PM

Dec 4 1PM

Dec 4 11PM

Dec 5 12AM

Dec 17 4PM

Dec 17 5PM

Dec 17 6PM

Dec 17 7PM

Dec 17 8PM

Dec 17 9PM

Dec 17 10PM

Input tokensOutput tokens

Model mix

Tokens and spend by model

6 models · all workflows

Tokens by model

input + output

Claude Opus 4.5654.2K tokens

1227 calls30.7% of total

gemini-3-pro-preview574.6K tokens

1508 calls27% of total

GPT-5.2420.6K tokens

1168 calls19.8% of total

Kimi-K2384.3K tokens

1232 calls18.1% of total

GPT-5.1-CODEX70.7K tokens

216 calls3.3% of total

Qwen3-Coder24.3K tokens

89 calls1.1% of total

Spend by model

estimated cost

Claude Opus 4.5$12.9

71.5% of total

GPT-5.2$3.39

18.9% of total

gemini-3-pro-preview$0.67

3.7% of total

Kimi-K2$0.59

3.3% of total

GPT-5.1-CODEX$0.46

2.5% of total

Qwen3-Coder$0.01

0.1% of total

Analyzed sample

Spend by workflow

5 workflows · sample-scoped

Rows below show the analyzed sample. The headline spend, tokens, and savings above are projected to the full source pool.

WorkflowRuns / moTokens (in / out)SpendSavingsTop issue

Flight Reservation Modificationderived-flight-reservation-modification11686.2K / 401.1K$2.36$1.57Outcome cohort gap Flight Cancellation Refundderived-flight-cancellation-refund8757.1K / 189.7K$1.13$0.67Outcome cohort gap Otherderived-other5261.6K / 151.9K$0.77$0.50Outcome cohort gap Flight Bookingderived-flight-booking2419K / 76.2K$0.65$0.44Outcome cohort gap Flight Delay Compensationderived-flight-delay-compensation2117.6K / 79.9K$0.39$0.23Outcome cohort gap

Opportunities

What's causing waste

25 opportunities · $11.3/mo projected

Tool misuse

55 of 116 runs fail at full cost. Adding early exit checks before heavy tool calls could cut waste fast.

$1.445 workflows

Outcome cohort gap

47% of runs fail or partially pass. Failed runs average 19.8 tool calls and 934 input tokens versus 9.9 calls and 565 tokens for successful runs.

$1.015 workflows

Flight Reservation Modification$0.49high riskhigh confidence Flight Cancellation Refund$0.17medium riskhigh confidence Other$0.16high riskhigh confidence Flight Booking$0.13high riskhigh confidence Flight Delay Compensation$0.06medium riskhigh confidence

Tool error loop

The agent retried a broken tool call 193 times across 2 threads without changing its approach, burning context and making no progress.

$0.545 workflows

Flight Reservation Modification$0.24high riskhigh confidence Flight Cancellation Refund$0.11high riskhigh confidence Other$0.08high riskhigh confidence Flight Booking$0.07high riskhigh confidence Flight Delay Compensation$0.04medium riskhigh confidence

Output contract mismatch

Users are fixing incomplete or incorrectly shaped responses from your assistant. This suggests the assistant isn't following its output rules consistently. A validator can catch these problems before users see them.

$0.425 workflows

Flight Reservation Modification$0.19medium riskhigh confidence Flight Cancellation Refund$0.09medium riskhigh confidence Other$0.06medium riskhigh confidence Flight Booking$0.05medium riskhigh confidence Flight Delay Compensation$0.03medium riskhigh confidence

Spend by agent

5 agents

derived-flight-reservation-modification44.5%

1 workflow

derived-flight-cancellation-refund21.3%

1 workflow

derived-other14.5%

1 workflow

derived-flight-booking12.3%

1 workflow

derived-flight-delay-compensation7.4%

1 workflow