Customize Bugbot PR Review Effort
TL;DR
Admins and Individual plan users can now set Bugbot's effort level to Default, High, or Custom. High effort allows for deeper reasoning to find more bugs, while Custom effort uses natural language instructions to determine review intensity.
What changed
On May 11, 2026 Cursor added effort controls for Bugbot PR reviews. Admins and Individual plan users can now choose Default, High, or Custom modes.
Default keeps the existing speed-focused behavior. High mode increases reasoning time and cost to surface more issues. Custom mode accepts natural language rules that decide when to apply higher effort.
Users must be on usage-based billing. Internal data shows Default finds 0.7 bugs per run on average while High reaches 0.95.
Why it matters
The change gives teams direct control over review depth versus spend. Solo builders and small teams can now match Bugbot intensity to the risk level of each pull request instead of accepting a single fixed setting.
It pressures competing AI coding tools to expose similar levers. The bet is that granular control will reduce wasted reviews on low-stakes changes while catching more problems on critical ones.
How to use it
Open the Bugbot dashboard at cursor.com/dashboard/bugbot and select the effort-levels section. Pick Default, High, or write custom instructions in natural language.
Confirm usage-based billing is active first. Changes apply immediately to new PR reviews.
Watch for
Track whether Custom rules actually reduce average cost per PR while maintaining bug detection rates. Watch for a follow-up release that adds per-repository defaults or automatic effort suggestions based on file changes.
Who this matters for
- Vibe Builders: Use Custom mode to set natural language rules that focus review intensity on high-risk UI changes.
- Developers: Toggle High effort for critical PRs to increase bug detection from 0.7 to 0.95 per run via deeper reasoning.
Harsh’s take
Cursor is moving away from the one size fits all approach to AI code review. By exposing effort levels, they are effectively letting users trade compute cost for accuracy on a per-PR basis. This is a pragmatic shift: not every minor CSS tweak needs a deep reasoning pass, but core logic definitely does.
The Custom mode is the real winner here. Using natural language to define review intensity allows teams to bake their specific risk tolerance into the tool without complex configuration files. It sets a high bar for competitors who still treat AI reviews as a black box.
If you are on usage-based billing, this is an immediate win for balancing your monthly API spend against code quality.
by Harsh Desai
About Cursor
View the full Cursor page →All Cursor updatesGo deeper
More from Cursor
- FeatureCursor adds cloud agent management to the Agents window
Cursor sets up cloud development environments in under 10 minutes, spins up isolated cloud subagents using /in-cloud, and hands off sessions between local and cloud.
- FeatureCursor introduces /automate skill for automating repetitive tasks
Cursor's new /automate skill creates automations from plain language. Workflows trigger via Slack emojis or GitHub events while cloud agents access virtual computers.
- FeatureCursor Adds New Customize Page for Managing Workflows Plugins and Rules
Cursor introduces a Customize page to manage plugins, skills, MCPs, and rules in one place. It includes a marketplace leaderboard, prebuilt plugin canvases, and repository import support.