Cursor Sets Auto-review as Default for New Agentic Coding Users

Cursor

Jun 13, 2026

Cursor now enables Auto-review by default for all new users. This feature uses a classifier subagent to evaluate agent actions in context, deciding whether to allow, block, or request approval. Internal evaluations show the classifier achieves 97% accuracy, helping to minimize approval fatigue while maintaining safety during autonomous coding tasks.

The classifier is highly accurate on clear-cut cases and degrades only near ambiguous edge cases
Precision Recall
All runs - 97.39% accurate
Allow 97.45% 99.22%
Block 97.18% 91.21% — Classifier performance metrics showing precision and recall scores for Allow and Block categories across all runs.

View the full update on cursor.com

Cursor

@cursor_ai2d ago

Auto-review is now the default for all new users. A classifier subagent reviews actions in context before deciding whether to allow, block, or ask for approval. Our evals show it's 97% accurate, with most misses near ambiguous edges. https://t.co/H4LthU8A3v

591.2k

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Keep reading

Cursor Auto-review Mode Uses Subagents to Cut Approval Fatigue

Cursor launched Auto-review, a new run mode that delegates tool call approvals to a specialized classifier subagent. By automating safety checks for terminal and API actions, the system allows coding agents to work longer without human interruption.

Cursor Publishes CursorBench, Its Internal Agentic Coding Evaluation Methodology

OpenAIMar 15

Cursor Publishes CursorBench, Its Internal Agentic Coding Evaluation Methodology

Cursor published CursorBench, its internal eval suite that scores models on real coding agent tasks from actual developer sessions. Public benchmarks struggle to differentiate frontier models reliably — CursorBench produces more separation where it matters most.