GitHub just turned Copilot pricing into an engineering operations topic.
On April 27, 2026, GitHub announced that all Copilot plans will transition to usage-based billing on June 1, 2026. That sounds like a finance update. It is not. It changes how teams think about coding-agent adoption, model selection, code review automation, and developer budgets.
If you lead engineering, platform, or developer productivity, this is the part that matters: GitHub Copilot is moving away from a mostly flat-fee mental model and toward a system where usage depth, model choice, and workflow design directly shape cost.
What changed on June 1, 2026
GitHub is replacing request-based pricing with a usage-based system built around GitHub AI Credits. Instead of treating advanced usage as a fixed bucket of premium requests, GitHub now prices Copilot based on token consumption, including input, output, and cached tokens.
That means heavier sessions cost more than lighter ones, and frontier models cost more than cheaper models. In other words, Copilot pricing now behaves much more like cloud infrastructure or model API billing.
For individuals, GitHub says monthly Copilot Pro includes $10 in monthly AI Credits and Copilot Pro+ includes $39 in monthly AI Credits. Monthly Pro and Pro+ subscribers automatically migrate to usage-based billing on June 1, 2026. Annual subscribers stay on their current annual plan until expiration, but GitHub is also changing model multipliers for those users starting June 1.
How Copilot pricing now works for teams
The bigger change for businesses is how GitHub is structuring organizational usage.
Copilot Business and Copilot Enterprise now use pooled GitHub AI Credits at the billing-entity level. GitHub documents the standard included amounts as 1,900 AI Credits per Business user per month and 3,900 AI Credits per Enterprise user per month. Existing customers get promotional higher pools through September 1, 2026: 3,000 credits per Business user and 7,000 credits per Enterprise user.
The pooled structure is important. It means power users can draw more from the shared pool while lighter users offset them. That makes rollout easier than purely individual caps, but it also means one small group of heavy coding-agent users can materially change total spend.
GitHub also lets teams control usage with budgets at the enterprise, organization, cost-center, and user level. That is a meaningful shift. Copilot is no longer just a seat you buy. It is a metered resource you govern.
Why code review is the detail teams should not miss
The easiest mistake in this rollout is assuming everything is now just AI Credits. It is not.
GitHub’s documentation says Copilot code review is billed in two ways. First, token consumption is billed in AI Credits. Second, the agentic infrastructure behind code review consumes GitHub Actions minutes. Starting June 1, 2026, that applies on GitHub-hosted runners.
This matters because Copilot code review is no longer just an AI feature layered on top of pull requests. Under the hood, GitHub uses Actions runners to power the agentic capabilities of code review, including full project context gathering and other workflow-heavy behaviors. If you use larger GitHub-hosted runners, rates go up further. If you use self-hosted runners, those runs do not consume GitHub Actions minutes.
That creates a new optimization problem for platform teams. You are no longer only choosing whether to enable code review. You are choosing which runner strategy, which review coverage, and which repositories deserve the most expensive agentic review path.
What remains unlimited and what does not
One helpful point in the change: code completions and next edit suggestions are still not billed in AI Credits. For paid plans, they remain unlimited.
The metered side applies to the more agentic and chat-heavy surfaces: Copilot Chat, Copilot CLI, Copilot cloud agent, Copilot Spaces, Spark, third-party coding agents, and code review. That distinction matters because it separates classic assistant usage from the workflows that behave more like autonomous or semi-autonomous software work.
Put simply, the more Copilot acts like an agent, the more its economics start to look like agent infrastructure.
What engineering leaders should do before the switch
GitHub is rolling out preview tools in early May 2026 so customers can compare current spend with projected spend under the new model. Teams should use that window seriously.
Review your highest-cost workflows. Identify where Copilot cloud agent, CLI sessions, and code review are used most heavily.
Segment model usage. Frontier models are expensive. Not every task needs them.
Check code review exposure. If code review becomes default on many repositories, Actions usage may become the surprise line item.
Set budgets early. Budget controls are now part of Copilot governance, not an optional admin detail.
Decide where self-hosted runners make sense. For some organizations, that may be the cleanest way to control the new code review cost pattern.
The broader lesson is that AI coding rollout now needs FinOps thinking. Seat count is no longer the whole story.
The business takeaway
GitHub’s pricing change is really a product signal. Copilot is becoming less like a flat developer add-on and more like a layered AI work system.
Once pricing follows tokens, context depth, model selection, and runner usage, teams have to think architecturally. Which tasks deserve expensive frontier reasoning? Which ones can use cheaper models? Which workflows should be automated by default, and which should stay human-led?
That is why this update matters. It is not only about what GitHub will charge on June 1, 2026. It is about how the economics of coding agents are getting closer to the actual amount of work they perform.
For teams serious about AI-assisted software delivery, pricing is now part of agent strategy.