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 and 1.5k output tokens.

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

A context window goes in, a summary comes out. See what survives and what the re-read bill looks like. The 575k-token window, 0.75% survival rate, and 28 reread steps are illustrative defaults—not a measured median. Replace them with observations from your own workload.

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 and the current burn multiple to see when the budget runs dry.

A basic single-metric HPA estimate: desired replicas equal current replicas times current utilization over target utilization, rounded up and clamped to the minimum and maximum. CPU utilization is relative to the pod CPU request. It assumes at least one replica. Other HPA behaviors are not modeled.

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.

Conservative p99-based pool sizing: peak RPS times p99 latency in seconds, times a safety factor, spread across pods. This heuristic differs from formal Little's Law, which uses mean time in system.

Minutes down times revenue per minute, and 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.