GitHub Copilot now shows the exact AI credits you're burning
The usage page reports per-cycle credits, not a percentage, as Copilot adds firewalled code review, a repo-metrics API, and Codex restores 272K context.
Copy markdown
See your real credit burn, not a percentage
GitHub Copilot Business and Enterprise seats now see the actual AI credits spent this billing cycle on the usage page — with an admin budget it shows used-vs-budgeted, and without one it shows your running total. If you keep hitting opaque premium-request limits, you can finally pace yourself against a real number.
Codex gets its 272K context window back
OpenAI's Codex 0.144.6 (Jul 18) fixed bundled instructions for GPT-5.6 Sol, Terra, and Luna and restored their full 272,000-token context inside Codex after a misconfiguration had clipped it. If your Codex agent felt like it was forgetting mid-task, upgrading the CLI is the fix.
Copilot code review moves behind the firewall
GitHub added custom firewall support, custom setup steps, and independent runner configuration to Copilot code review (Jul 17), so teams on locked-down networks can run automated reviews. GitHub Mobile also gained “Fix with Copilot” straight from a review comment.
A usage API to chart your own Copilot adoption
Two new endpoints (GA Jul 17) return a daily, per-repository breakdown of pull-request activity for Copilot coding agent and code review. Build your own adoption dashboards instead of guessing how heavily your team leans on the agent.