Review load calculator for parallel coding agents
One reviewer can keep up with as many agents as their daily review minutes divided by the review minutes one agent adds, with second rounds counted as extra reviews. At the defaults (4 agents, 3 pull requests each, 20 minutes per review, 25 percent rework, 4 reviewer hours), that figure is 3.2 agents. Four agents need 5 hours of review against 4 available.
Reviewed November 2026
- Review minutes needed per day
- minutes
- Review hours needed per day
- hours
- Agents one reviewer can keep up with
- agents
What does this calculator work out?
It works out three numbers. The first is the review minutes needed per day. The second is that figure in hours. The third is the number of agents one reviewer can keep up with at those rates. Agents write code in parallel, but review is a single queue that one person works through in order.
A pull request sent back for a second round costs another review, so the rework share raises the load without adding new work. The result is arithmetic on your own numbers and says nothing about any product.
What do you enter?
- Agents running in parallel (agents) How many agents open pull requests against the same codebase on a normal day. Default 4, allowed range 1 to 50.
- Pull requests per agent per day (pull requests) Count the pull requests one agent opens in a day, using your merged and closed history for a typical week. Default 3, allowed range 1 to 50.
- Minutes to review one pull request (minutes) The time from opening the diff to a decision, including reading tests and running the branch if you do. Default 20, allowed range 1 to 240.
- Share sent back for a second round (percent) The percentage of pull requests that return after requested changes. Default 25, allowed range 0 to 100.
- Reviewer hours available per day (hours) Hours one reviewer can actually spend on review, not hours at the desk. Default 4, allowed range 0.5 to 16, in steps of 0.5.
How is the result calculated?
Start with the pull requests all agents open in a day. Multiply by one plus the rework share, so a 25 percent rework rate means each pull request counts as 1.25 reviews on average. Multiply the resulting reviews by minutes per review to get daily review minutes, then divide by 60 for hours.
For the agent count, work out the review minutes one agent adds per day and divide the reviewer's available minutes by that figure.
reviewsPerDay = agents * pullRequests * (1 + percent / 100) minutesNeeded = reviewsPerDay * minutes hoursNeeded = minutesNeeded / 60 minutesPerAgent = pullRequests * (1 + percent / 100) * minutes agentsSupported = (hours * 60) / minutesPerAgent
Worked example
With the defaults, 4 agents opening 3 pull requests each produce 12 pull requests a day. A 25 percent rework share makes that 15 reviews. At 20 minutes each, the day needs 300 review minutes, which is 5.0 hours. One agent adds 75 minutes a day, and 4 reviewer hours is 240 minutes, so one reviewer keeps up with 3.2 agents. Four agents leave 60 minutes of review uncovered each day.
What does the result not tell you?
The model is illustrative arithmetic on the numbers you enter. It assumes every review takes the same time, though a small fix and a large refactor rarely do, and a second round is often shorter than the first. Adjust the minutes input to approximate that.
It also treats reviewer time as the only constraint. Merge conflicts, waits on CI and context switching are left out, and agents that touch the same files create work the formula does not count. For ways to keep their changes apart, see parallel agents with git worktrees.
Some platforms add review rules of their own. On GitHub, if your repository requires pull request approvals, your approval of a Copilot pull request does not count toward the required number, and another reviewer must approve it before it can be merged. Where that rule applies, plan for a second reviewer on those pull requests.
To shorten the queue itself, read the guide to the coding agent review loop.
What else do readers ask?
What if the result says my reviewer supports fewer than one agent?
Your review time per pull request or your pull request volume is higher than one person's hours can cover. Lower the pull requests per agent, add reviewers, or ask the agents for smaller diffs.
Should I count automated checks as review time?
Count only the minutes a person spends. Time a pipeline spends running does not use reviewer hours, though waiting on it can stretch your day.
Why does the rework share matter so much?
Each returned pull request is read again, so a high rework rate adds reviews without adding any new pull requests. Moving it from 25 to 50 percent in the calculator shows the effect directly.