Free Engineering Calculators

Price one agent turn from context mix, then scale to a month. Starts from a heavy agent session: 80k input tokens at 87.5 percent cache hits plus 1.5k output tokens.

Turn events per second into a monthly envelope: edge rate times seconds per month at a blended price, plus fixed platform spend.

A context window goes in, a summary comes out. See what survives and what the re-read bill looks like. Defaults mirror the measured median: 575k crushed to about 4.3k, then roughly 28 steps of re-reading.

Failed agent attempts rebill every retry. Multiply tasks by retries by attempt cost and watch the storm's share of the budget.

Turn an SLO into minutes. A 99.9 percent objective leaves 43.2 minutes a month. Enter what is already burned plus the current burn multiple to see when the budget runs dry.

The real Kubernetes autoscaler formula. Desired replicas equal current replicas times current utilization over target utilization, rounded up. Defaults scale 8 replicas at 72 percent toward a 60 percent target.

Spans per second times bytes per span is your daily ingest. Price it, then compute the keep rate that fits the budget. Defaults ingest 50k spans a second at 2 KB against a $20k monthly budget.

Little's law for pools. Concurrent connections equal peak RPS times p99 latency in seconds, times a safety factor, spread across pods. Defaults mirror a 10k TPS payments switch at 45ms p99.

Minutes down times revenue per minute, plus the SLA credit on the cloud bill. Assumes a 43,800-minute month. Defaults model a four-hour outage against $4M monthly revenue at risk.