Cursor will loop to read and analyze the same file and cannot stop

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

cursor will loop to read and analyze the same file and cannot stop, and the token consumption is particularly fast,Problems that occurred after updating to the latest version

Steps to Reproduce

This happens every few times during normal use.

Operating System

Windows 10/11

Version Information

Version: 2.4.37 (system setup)
VSCode Version: 1.105.1
Commit: 7b9c34466f5c119e93c3e654bb80fe9306b6cc70
Date: 2026-02-12T23:15:35.107Z
Build Type: Stable
Release Track: Default
Electron: 39.2.7
Chromium: 142.0.7444.235
Node.js: 22.21.1
V8: 14.2.231.21-electron.0
OS: Windows_NT x64 10.0.26100

For AI issues: which model did you use?

composer 1

For AI issues: add Request ID with privacy disabled

ab1baf68-bd3a-49bf-9c34-6b0ab75fbb47

Does this stop you from using Cursor

No - Cursor works, but with this issue

Hey, thanks for the report.

From the screenshot, it looks like the agent gets stuck in a loop and keeps reading FileBrowserViewModel.kt over and over without moving forward. This is a known issue that can happen with some models and prompts.

Here are a few things to try:

  • Try a different model. Switch from composer-1 to Claude Sonnet 4.5 or GPT-5.2 and see if the loop still happens.
  • Stop and rephrase. When you notice the loop starting, stop generation and try breaking the task into smaller steps.
  • Start a new chat. If the loop starts, open a new chat instead of continuing the current one.

Question: does this happen consistently with the same prompt, or is it random? Also, are you using Agent mode or Ask mode?

Let me know how it goes.

it is random, both agent and ask, I’ve tried starting a new chat or different model, but the problem still occurs randomly.

If this is happening randomly and in different models and modes, it’s not just a quirk of one specific model.

I’ve shared your request ID with the team so they can investigate. A couple of quick questions to narrow it down:

  1. When the looping happens on a model other than composer-1, which model is it exactly? (Claude, GPT, etc.)
  2. About how big is the FileBrowserViewModel.kt file? Very large files can sometimes trigger this behavior.

For now, the most reliable way to stop it is to click Stop as soon as you see the looping start. That should at least limit token usage.

composer1.5 and I tried auto mode, FileBrowserViewModel.kt is about 15kb


95e13fb4-54b3-4157-9aac-b0923955d67b
It appeared again