GitHub Copilot usage limits, planned around instead of discovered

The meter is a credit allowance with a monthly reset. A month that ignores the clock always ends the same way.

·3 min read

GitHub Copilot usage limits are easier to plan around once you stop treating them as a request count. On current monthly plans the meter is GitHub AI Credits, one credit worth $0.01, and the allowance refills at 00:00 UTC on the first day of the calendar month. A month of work that ignores that clock will always feel like a surprise.

What the limit is now

Each paid plan includes a base credit amount and a flex allotment on top. Base credits are spent first; the flex part is applied after them, at the same model rates, across the editor, GitHub.com and the CLI. Completions and next edit suggestions sit outside the meter on paid plans, so the work that runs out first is chat, agents and code review.

Work Metered in credits What drives the cost
Inline completions No Nothing, on a paid plan
Chat Yes Model choice and conversation length
Agent session Yes Many model calls inside one task
Code review Yes The effort level you set
Third-party agents Yes Preview features, same meter

Put the reset on the calendar

The reset is fixed to the calendar month, not your signup date, and unused credits do not roll over. That has a practical consequence for anyone outside UTC: the reset lands in the middle of a working day, or on the last afternoon of the previous month, depending on the offset. A heavy migration started two hours before the refill can either finish on the new allowance or die on the old one. Check the clock before the push, not after.

Plan in weeks, not in totals

A plan allowance is a monthly number, but work arrives in weeks. Split the cycle into four windows and give each one a job.

  • Week one: normal work, and a note of the daily burn.
  • Week two: the heavy refactors and migrations, while the allowance is fresh.
  • Week three: review, documentation and smaller fixes.
  • Week four: protect what is left for the unplanned bug.

That is not a budget in credits. It is a budget in attention, and it works because most teams already know which week the release lands.

When the allowance runs low

GitHub gives three routes: upgrade, set a budget for additional usage and pay for what you use, or wait for the next cycle. An upgrade during a cycle is charged as the difference between plans, and existing usage counts inside the larger allowance. For a team, additional usage is governed by budgets set at the user, organization or enterprise level, so the person who can raise the ceiling is usually an admin, not the developer who hit the wall.

Do not plan from someone else's number

Published allowances go stale. Flex allotments change, model prices change, and a plan that suited a team in June may not fit in November. The only number worth planning around is your own, measured across two cycles at the same point in each. UsageAtlas keeps the reported Copilot balance and reset on your desktop, free and open source, so the second cycle has something to compare against. It is a tracker, not a billing record, and the official figure stays on GitHub.

Questions

Do Copilot credits roll over?

No. GitHub's documentation says the included allowance resets at 00:00 UTC on the first of each month, and unused credits are forfeited.

Are code completions counted against the limit?

No. Completions and next edit suggestions are not billed in AI credits and stay unlimited on paid plans, though rate limits still apply.

What happens when a team runs out?

The organization can add budget for additional usage, or the member can wait for the reset. Business and Enterprise allowances are pooled, so one heavy seat affects the shared pool.