Cap GitHub Copilot credit pools before one team burns the budget

GitHub has added AI credit pools for cost centers. Confidence level: confirmed. Enterprise owners can now cap how much of the monthly included Copilot AI credit pool a cost center can draw, with REST API support available now and settings UI support still pending.

Caption: GitHub's official changelog social image for new release notes. Source: GitHub Changelog.

What changed

GitHub published the cost-center AI credit pool update on July 2, 2026. The feature is for Copilot Business and Copilot Enterprise on GitHub Enterprise Cloud. It lets enterprises stop one cost center from consuming more included AI credits than the Copilot licenses assigned to that cost center fund.

This is not the same as a normal budget. GitHub says an AI credit pool governs the included monthly Copilot AI credits before additional usage starts. A cost-center budget governs metered usage after the included pool is exhausted. Enterprises can use both controls on the same cost center.

Key takeaways

• AI credit pools are available through the REST API today.
• Cost-center settings UI management is still coming later.
• The cap is calculated from assigned Copilot Business and Copilot Enterprise licenses.
• Enterprises can choose whether a cost center is blocked at the cap or allowed to continue into additional spend when overages are allowed.
• The feature fits chargeback, department budgets, and agent-spend governance.

Availability and access

GitHub says the feature is available for Copilot Business and Copilot Enterprise on GitHub Enterprise Cloud. It supports the earlier enterprise teams and cost-center user-level budget releases, which together form GitHub's first suite of AI usage spend controls at scale.

Admins should check whether their enterprise already uses cost centers, whether teams or users are assigned correctly, and whether the REST API route fits their billing workflow. If finance teams need self-service controls, wait for the UI or build a small internal approval flow around the API.

Practical LinkLoot angle

Copilot agents, subagents, and model-heavy workflows make shared AI credits harder to govern. Without pool boundaries, a fast-moving engineering group can spend credits funded by another group's licenses before anyone sees the chargeback mismatch.

The practical workflow is to map cost centers to teams or departments, enable AI credit pools where chargeback matters, and pair the pool with budgets for metered overage. Teams running agent workflows should also keep per-session limits in scripts and CLI runs. For more stack-level planning, see LinkLoot's guide to AI workflow automation.

What to verify before you act

• Confirm your organization uses GitHub Enterprise Cloud with Copilot Business or Copilot Enterprise.
• Check whether cost centers are mapped to users, teams, departments, or client work in a way finance can defend.
• Review REST API support before promising nontechnical admins a UI workflow.
• Decide whether a capped cost center should block usage or continue into additional spend.
• Compare AI credit pools with budgets and session limits so each control has a clear owner.

Source check

Confirmed by: GitHub's July 2, 2026 changelog entry and GitHub billing documentation for cost centers and usage-based budgets. Those sources establish the feature, the REST API-first availability, the difference between included credit pools and cost-center budgets, and the intended enterprise billing context.

Independent context: Changes.Watch tracks the same GitHub release note and summarizes the AI credit pool support as a cost-center control for monthly included Copilot AI credit usage.