Agent Capacity Calculator
Estimate how many builds a Jenkins agent fleet can complete per day given agent count, working hours, average build time, and utilization.
Inputs
Total build agents in the fleet.
Hours per day the fleet is available to run builds.
Average time a build occupies one agent.
Fraction of theoretical capacity actually used (accounts for idle time, queueing overhead, failed builds).
Estimated Builds per Day
1,008.0builds
Builds per Agent per Day
100.80builds
Theoretical Max (100% Utilization)
1,440.0builds
Step by step
Theoretical builds/day = agents × (hours ÷ avg build time)
10 × (24hr ÷ 0.167hr)
= 1440.0 builds
Actual builds/day = theoretical × utilization
1440.0 × 70%
= 1008.0 builds
How it works
Each agent's theoretical daily capacity is its available hours divided by the average build duration; multiplying by agent count gives the fleet's theoretical maximum. Real fleets never hit 100% utilization — time is lost to provisioning, queueing gaps, retries, and idle periods — so multiplying by a realistic utilization factor (commonly 60–80% for busy CI fleets) converts the theoretical ceiling into an achievable daily throughput estimate.
Formula
buildsPerDay = agents × (hoursPerDay / avgBuildTimeHrs) × utilization
- A
- Number of agents
- H
- Available hours per day
- t_b
- Average build time in hours
- u
- Utilization factor (0 to 1)
- B
- Estimated builds per day
Frequently Asked Questions
What utilization rate should I use?
70% is a reasonable default for actively used CI fleets; ephemeral cloud agents that spin up on-demand can approach 85–90%, while static agent pools with uneven load across timezones or teams often sit closer to 50–60%.
How does this differ from the executor requirement calculator?
This calculator estimates achievable throughput from a fixed fleet size; the executor requirement calculator works in the opposite direction, solving for how many executors are needed to hit a target throughput.
Does agent count here mean executors or physical/VM agents?
It represents build agents each assumed to run one build at a time; if agents run multiple executors concurrently, multiply agent count by executors-per-agent first to get an equivalent single-build-slot count.