Cursor has rolled out a major update to Bugbot, its automated code review tool, saying the system now runs more than three times faster, costs 22% less, and identifies 10% more bugs in each review.
The company said the changes are already having a noticeable effect on review times. According to Cursor, 90% of Bugbot runs now complete in under three minutes. The faster turnaround is meant to help developers spot problems earlier in the workflow and move code through review more quickly.
Bugbot is designed to analyze code changes and flag potential issues before code is merged. Cursor is positioning the latest update as both a productivity improvement and a cost reduction for teams that rely on automated reviews at scale. The company said the combination of speed, lower expense, and better bug detection should help developers catch issues sooner and reduce the time spent waiting on feedback.
Alongside the performance update, Cursor introduced a way to use Bugbot and its Security Review agent before code is pushed. Developers can now use the /review command to choose which agents to run, or call /review-bugbot and /review-security directly.
Cursor says the new workflow is intended to catch problems earlier, before a pull request is opened or code is sent to GitHub or GitLab. The company also said /review can sync with Bugbot on those platforms. If a developer runs a review locally and then opens a pull request with the same diff, Bugbot can recognize the duplicate work, skip the second review, and leave a note that the code has already been checked.
The feature is available in Cursor 3.7 and later, as well as on cursor.com/agents. Support for the CLI is planned for later.
Cursor also said users can now configure Bugbot so that it reviews only the code that changed since the last check, rather than reprocessing an entire pull request every time new commits are added. The company said this should keep feedback more focused on the latest changes and avoid repeated flags on code that had already been examined.
That adjustment may be especially useful for active pull requests that receive multiple rounds of updates. By limiting reviews to new changes, Cursor is aiming to reduce noise while keeping the tool aligned with the most recent edits.
Cursor attributed the latest performance improvements to upgrades in its underlying review system and to progress in training Composer 2.5, the model now powering Bugbot. The company said its model work is one of the main ways it plans to keep improving Bugbot over time.
Cursor also noted that Bugbot respects model block lists. If an organization has opted out of Composer 2.5, the tool will fall back to another available model. In those cases, the company said performance and speed may vary depending on the configuration.
The update adds to Cursor's broader push to make Bugbot a faster and more efficient part of the development process. For teams using automated code review, the company is betting that quicker feedback and lower costs will make it easier to identify issues without slowing down code shipping.