Agent Planning, Better Context & Faster Tab

Source: Cursor Date: 2025-07-03 URL: https://cursor.com/changelog/1-2

Summary

Cursor 1.2 ships agent structured to-do lists with task dependencies visible in chat and Slack, message queuing for sequential follow-ups, PR indexing and semantic search across PR history and GitHub comments, automatic merge conflict resolution via agent, and Memories graduating to GA with user approval workflows. Tab completions ~100ms faster, first-token latency down 30%. Background Agent gains automatic PR template compliance and auto-pull.

Implications

Agent-IDE feature race. PR indexing is a semantic search expansion that gives agents historical PR context — not just current code. An agent that can search PRs for “how did we handle X before” is substantially smarter about project conventions than one that only reads the current tree. Combined with merge conflict resolution, Cursor is positioning the agent as a collaborator in the full Git workflow, not just a code generator.

Agent-IDE feature race (Memories GA). Memories graduating with user approval workflows signals Cursor learned from the beta: users need to control what agents remember, not just benefit from automated persistence. The approval step is a trust mechanism — it makes Memories safe for team deployments where one user’s preferences shouldn’t pollute shared agent behavior.

Watch: Whether PR semantic search integrates with Bugbot’s rule learning (past PR review patterns informing future automated reviews) and whether merge conflict resolution extends to more complex 3-way conflicts.