Agent mode cannot be completed smoothly, and sometimes it gets interrupted

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor CLI

Describe the Bug

When running, it cannot execute smoothly and often gets interrupted, requiring me to repeatedly say: “Please continue.” Sometimes the code finishes executing but isn’t recognized. Other times, I’m halfway through writing a document when it hasn’t been saved, and when I say “Please continue,” it restarts the document writing process.

Steps to Reproduce

When writing code in agent mode, it’s often difficult to complete smoothly and frequently gets interrupted.

Operating System

Windows 10/11

Version Information

Version: 2.4.27 (user setup)
VSCode Version: 1.105.1
Commit: 4f2b772756b8f609e1354b3063de282ccbe7a690
Date: 2026-01-31T21:24:58.143Z
Build Type: Stable
Release Track: Early Access
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 10.0.26200

For AI issues: which model did you use?

I have been consistently using agent mode with vibe coding. I’ve said this many times recently: please continue. The reason is that it always interrupts execution and cannot complete automatically. Previously, I might have used 1.5 ultra packages per month, but in the last month, I used 3 ultra packages. This is because it frequently interrupts, and I have to keep saying: please continue.

For AI issues: add Request ID with privacy disabled

41d35f32-a244-4fa8-bb49-a836e4d2a3c5

Does this stop you from using Cursor

No - Cursor works, but with this issue

5

Hey, thanks for the report. I can see this is really getting in the way of your work, and I get how frustrating the 3x token usage is.

I’ve seen similar reports where the agent stops in the middle of a task. To dig in, we’ll need a bit more info:

Diagnostics:

  • Screenshots of the moment the agent stops, especially if you can see the “thought process” where it plans to continue but then stops
  • Console errors: Help > Toggle Developer Tools > Console tab (if anything shows up there when it stops)
  • Cursor Settings > Network > Run Diagnostics, please share the results

A few questions:

  • Does this happen on every task or only certain types, like big files, lots of files, or specific operations?
  • Does it happen in different projects too?
  • When you say “code finishes executing but isn’t recognized”, do the commands run in the terminal but the agent doesn’t notice them?