GitHub Copilot weekly releases — September 14

Code review in GitHub Copilot now resolves comments you've already addressed in a later review while leaving unresolved feedback open, and it suggests a commit message when you apply one of its suggestions. Reviews can also run shell tools to validate changes, and Lite reviews now combine findings from multiple agents.

• Auto model selection adds three tiers, efficiency, balance, and intelligence, for weighing cost, quality, and response time; all three pick from the same available models, and the change is rolling out in VS Code, Copilot CLI, and the Copilot app.
• Business and Enterprise admins get generally available usage metrics in VS Code's Agents window: daily active users, session counts, and user message totals, separate from the existing editor-window metrics.
• Copilot can now suggest values for repository custom properties, for example "yes" and "no" for an internet-facing property, in public preview; org and enterprise owners can turn this on or off via a policy.
• Organizations on usage-based billing can request a higher Copilot budget when they hit their AI credit limit, with owners and billing managers able to approve, adjust, or deny requests; this is now generally available for Business and Enterprise plans, except enterprises with managed users.
• The Copilot app adds a Sentry canvas: review a crash report's errors and stack trace, then work with Copilot to investigate the cause, validate a fix, and prepare a pull request.
• In VS Code 1.138, agents in the Agents window can run with a project's own tools and dependencies via local Dev Containers, rolling out gradually and requiring Docker plus a supported Dev Container configuration.

GitHub

Where the world builds software — Git hosting, pull requests, issues, Actions CI/CD, and Copilot, free for public and private repos.

View GitHub →

Read the original announcement →