Cursor: Bugbot Autofix

Cursor

Cursor has launched Bugbot Autofix, bringing autonomous bug-fixing capabilities to its AI-powered code review agent. When Bugbot identifies issues in a pull request, it now spawns cloud agents that run in isolated virtual machines to test proposed fixes before submitting them directly to the PR. Over 35% of Bugbot Autofix changes are merged into the base PR, and since launch the tool’s overall resolution rate has climbed from 52% to 76%. The feature is now generally available to all Bugbot users and can be configured for either manual approval or fully automated push to branch.


Bugbot Gets the Ability to Fix What It Finds

Cursor has expanded its Bugbot code review agent with a significant new capability: Autofix. Rather than simply identifying bugs and leaving remediation to the developer, Bugbot can now automatically propose and test fixes directly on pull requests — closing the loop between detection and resolution that has historically required human intervention.

How Autofix Works

When Bugbot reviews a pull request and identifies an issue, Autofix triggers a cloud agent that spins up in its own isolated virtual machine. The agent tests the proposed fix within that environment, verifying the change before surfacing it to the developer. This is not static patch suggestion — Bugbot's agents actively run the software they are modifying, giving the fix a meaningful degree of validation before any PR comment is posted.

Once the agent completes its work, Bugbot posts a comment on the original PR containing a preview of the proposed changes. Developers can merge these changes using a provided @cursor command. Alternatively, teams can configure Autofix to push changes directly to the branch with no interaction required — a fully automated mode suited for workflows where speed and throughput take priority.

Performance and Adoption

The numbers behind Bugbot Autofix are notable. Over 35% of Autofix-proposed changes are merged into the base PR by developers — a strong signal of trust in the quality of the automated fixes. More broadly, Bugbot's overall resolution rate has climbed from 52% at launch to 76% over six months of iteration, while the average number of issues identified per run has nearly doubled. Cursor processes over 2 million pull requests monthly through Bugbot, with enterprise customers including Rippling, Discord, and Samsara among its active users.

Enabling Autofix

Autofix is now generally available to all Bugbot users, having exited beta. To enable it, users navigate to their Bugbot dashboard at cursor.com/dashboard?tab=bugbot. Full documentation, including configuration options for the BUGBOT.md file that lets teams define codebase-specific review rules, is available at cursor.com/docs/bugbot#autofix.

What's Next for Bugbot

Cursor has outlined several directions for Bugbot's continued development. Future capabilities include allowing Bugbot to verify its own findings by running code to confirm a bug actually manifests, enabling deep research mode for complex or ambiguous issues, and introducing custom automation workflows so teams can extend Bugbot beyond code review into other parts of their development pipeline. Cursor is also working toward continuous codebase scanning — shifting Bugbot from a reactive PR tool to a proactive quality monitor.