Cursor Bugbot Lets You Tune PR Review Effort for Faster or Deeper Finds

TL;DR

• Cursor's AI code reviewer, Bugbot, now allows configurable effort levels for Pull Request (PR) reviews.
• Users can choose between Default (speed-optimized), High (deeper analysis), or Custom (natural language instructions).
• The High effort level can find more bugs, averaging 0.95 bugs per run compared to 0.7 with default.
• Customization requires being on usage-based billing and is available for Team admins and Individual plan users.

Cursor is rolling out a significant update to its AI-powered code reviewer, Bugbot, introducing customizable effort levels for Pull Request (PR) reviews. This new feature empowers teams and individual developers to fine-tune how intensely Bugbot analyzes their code, striking a balance between review speed and the thoroughness of bug detection. This flexibility ensures that the AI's assistance can be tailored to specific project needs and timelines, making the code review process more efficient and effective.

The core of this update lies in three distinct configurations for Bugbot's effort. The Default setting maintains the current behavior, prioritizing quick turnaround times for reviews. For those situations where a more exhaustive check is required, the High effort level allows Bugbot to dedicate more reasoning time, potentially uncovering a greater number of issues. Finally, the Custom option offers unparalleled flexibility by enabling users to describe in natural language when they want Bugbot to engage its default or high effort modes, with Cursor dynamically adjusting the AI's approach based on these instructions.

This enhancement to Bugbot's capabilities directly addresses a common challenge in software development: the trade-off between speed and quality in code reviews. By offering configurable effort levels, Cursor allows teams to adapt their AI reviewer to different stages of development or varying levels of code complexity. For instance, a quick feature branch might benefit from a default, fast review, while a critical security patch could warrant the deeper analysis of a high-effort review to catch potential vulnerabilities.

The data shared highlights the tangible impact of these effort levels. With the default setting, Bugbot identifies an average of 0.7 bugs per run, with a commendable 79% resolution rate before merging. Upping the ante to high effort pushes this average to 0.95 bugs per run, suggesting a noticeable increase in the quantity of issues surfaced. This means teams can proactively address more potential problems, leading to more robust and reliable codebases.

To take advantage of Bugbot's new customizable effort levels, users must be on a usage-based billing plan. This requirement ensures that the system can accurately track and charge for the increased computational resources utilized during higher-effort reviews. Once this prerequisite is met, Team administrators and users on an Individual plan can access these settings.

Getting started is straightforward. More detailed information and guidance can be found in the official documentation, which elaborates on the nuances of each effort level and how to configure custom instructions. For direct access to manage these settings, users can navigate to their Bugbot dashboard. This allows for quick and easy implementation, enabling teams to immediately begin optimizing their AI-assisted code reviews.

Summary

• Cursor's Bugbot now features adjustable effort levels for PR reviews: Default, High, and Custom.
• High effort can identify more bugs, averaging 0.95 per run compared to 0.7 with default.
• Users need to be on usage-based billing to customize these effort levels.
• Configuration is available for Team admins and Individual plan users via the Bugbot dashboard.

Source: Bugbot Effort Levels · Cursor