GitHub's $10-a-month Copilot Pro plan now comes with 1,500 AI credits. Each credit costs one cent, so that is $15 worth of usage. Typing with code suggestions is free on paid plans. Handing whole tasks to an AI agent is what uses up credits.
What changed
Third-party trackers say GitHub swapped its old premium-request quotas for credits on June 1, 2026. What is new this fall is what sits on top. Copilot in VS Code, a popular code editor, now runs three coding agents side by side: GitHub's own, Anthropic's Claude and OpenAI's Codex. All of them draw from the same credit meter.
What uses credits
- Chat, agents, Copilot CLI and Spaces use credits.
- Code review uses credits, and since June 1 it also uses GitHub Actions minutes.
- Opt-in secret checks and the security-review command use credits in public preview.
- Code completions and next-edit suggestions do not use credits on paid plans.
- Local sandboxing is free, with no additional charge.
One independent breakdown reports that charges follow the text a model reads and writes, including cached text. A long agent session reads far more than a short chat. Cached reads cost less than fresh ones, so how a session is set up affects the bill as much as which model runs it.
How the plans compare
The more a plan costs, the more usage each dollar buys. Pro gives $15 of credits for $10, about 1.5 times the price. Pro+ gives $70 for $39, roughly 1.8 times. Max gives $200 for $100, a clean 2 times. GitHub also lists Free, Business, Enterprise and a Student tier. It does not publish an allowance figure for Free and Student on the pages it serves today.
Newer premium models tend to reach Pro+ and above first. Pro users get Claude Sonnet 5.5 and Haiku 5.5, but not Claude Opus 5.5.
What to do
Teams should decide which plan fits how they work. Developers who mostly accept inline suggestions will barely touch their allowance. Those who run agents for hours will not. In Copilot settings, you can see usage and the reset date, and set a budget with alerts at 75%, 90% and 100%. A $10 overage budget buys 1,000 credits.
Two newer features, agent automations and Agent Merge, are still in preview. Automations run recurring work on a schedule. Agent Merge handles review feedback and failed checks. GitHub labels both as previews, so treat them as experiments. An unattended overnight session is the clearest way to run up a surprise bill.
The takeawayPick a Copilot plan based on how much your team uses agents, and set credit budget alerts before letting any agent run unattended.