GitHub Copilot adds new model selection, code review features, and Sentry integration

GitHub Copilot now offers 'efficiency,' 'balance,' and 'intelligence' tiers for auto model selection, enabling users to prioritize cost, quality, or response time. Code review functionality has been improved to resolve addressed comments and suggest commit messages, alongside support for shell tools and combined findings from multiple agents in Lite reviews.

For Business and Enterprise users, new usage metrics are generally available in the VS Code Agents window, providing insights into adoption and activity. Copilot also now offers suggestions for repository custom property values in public preview, and a budget request system for AI credits is generally available for usage-based billing plans.

The GitHub Copilot app now integrates with Sentry, allowing users to move from crash reports to code fixes. This new Sentry canvas enables reviewing errors, stack traces, and context, then collaborating with Copilot to investigate, validate fixes, and prepare pull requests.

VS Code Agent users can now run agents with project tools and dependencies using local Dev Containers. An opt-in preview feature automatically marks inactive sessions as 'Done' once all associated pull requests merge, with optional deletion. Users can also create pull requests directly from Agent Host sessions within the Agents window.