Skip to content

Cursor Bugbot: self-improving rules, MCP support, and higher fix rates

Cursor's Bugbot can now learn rules in real time, connect to MCP servers, and ships its highest bug resolution rate yet.

Cursor’s latest Bugbot update adds real-time self-improvement, MCP support, better Autofix, and the highest resolution rate the tool has recorded.

The self-improvement piece is the most significant change. Bugbot can now learn rules on the fly — meaning it adjusts its behavior based on what it encounters during a session rather than relying solely on static configuration. That closes a gap where edge cases or project-specific patterns would slip through because the bot had no way to update its own understanding mid-run.

MCP support extends Bugbot’s reach. With Model Context Protocol integration, Bugbot can pull context from external tools and services during analysis — the same pattern that’s made MCP useful across other AI coding workflows. If your project connects to databases, APIs, or internal tooling via MCP, Bugbot can now factor that context into its bug detection.

Autofix also got improvements, though Cursor hasn’t published specific metrics on what changed there. What they did call out: this release hits the highest resolution rate Bugbot has achieved. That’s a meaningful signal — resolution rate is the number that actually matters when evaluating an automated bug-fixing tool.

Practical takeaway: If you’re already using Bugbot, pull the latest Cursor update and run it on a project with MCP servers configured. The combination of real-time rule learning and external context access is where the new capability shows up most clearly.