GitHub Copilot moves to token‑based AI Credits on June 1, 2026

GitHub changed how it charges for Copilot on June 1, 2026, replacing its premium‑request unit model with a usage‑based system called GitHub AI Credits.

Under the new system every Copilot plan includes a monthly allowance of AI Credits and additional use is billed by token consumption. GitHub defines one AI credit as $0.01 USD.

Credits are consumed based on tokens — input, output, and cached tokens — and are converted from model token rates into AI Credits at published per‑model prices. That finer granularity is designed to charge more precisely for high‑context and long agent sessions.

GitHub also kept monthly subscription prices the same but tied each plan’s included credits to its dollar price, and it is offering pooled credits for organizations so teams can share allowances. The company provided preview bills in accounts ahead of the switch to help admins estimate impact.

The shift matters most for agentic workflows and long, high‑context sessions that keep large amounts of context or call multiple models. GitHub’s Copilot agent features, Copilot CLI sessions, and cloud agents can consume many more tokens than short IDE completions, so those workflows will be metered more finely under AI Credits.

Developers and teams reported sharp increases in projected bills for heavy agent use after the change went live, with some public reports and forums flagging 10x to 50x jumps for certain workloads. That early backlash underscores how usage‑based billing changes developer economics for power users.

GitHub framed the move as a response to rising inference costs and to better align pricing with actual compute consumption. Company posts and briefings said the flat premium‑request model no longer matched the variety of ways customers run Copilot today.

Admins get tools to manage the transition: billing overview previews, pooled organization credits, and budget controls that can alert or pause charges when thresholds are reached. GitHub’s documentation emphasizes setting budgets and reviewing preview bills to avoid surprises.

Some product changes accompanying the billing rollout affect where costs appear. GitHub says Copilot code review now consumes GitHub Actions minutes in addition to AI Credits, which shifts parts of review cost into existing Actions billing lines.

The timing of the switch coincided with broader Build‑era product updates at Microsoft and GitHub that emphasize agent platforms, Copilot Studio, and tighter model routing for developers. Those announcements illustrate why vendors are pressing to make agent use commercially sensible and instrumented.

Practical steps for developers: run the preview bill in your billing overview, enable budgets and alerts, audit long agent sessions, and consider model selection and caching strategies to reduce token counts. For teams, pooled credits and the offered short‑term bonus credits for some business plans may smooth the initial months.

The big picture is that GitHub has aligned Copilot billing with tokenized API economics used across cloud AI services. That alignment promises more accurate cost signals for providers, but it will force many heavy users to redesign agent workflows or accept higher bills as token‑heavy capabilities become central to development. Watch GitHub’s docs and billing pages for further tweaks.