My Cursor's terminal tool no longer runs commands properly

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I’m working on an application in Ruby On Rails using the WSL as my main terminal. One of the bash scripts I’m trying to run is “bin/cucumber” which essentially is the bundle exec cucumber command with a flag option translating to a specific flow. In my wsl terminal, I can execute it no problem and see an output that is consistent with cucumber execution. In the chat interface, the Agent is unable see the output executed. I’ve seen other posts that have similar issues with the Agent in WSL and I’m unable to diagnose the issue. I’ve lost over $40 in tokens in the last month because of this issue. A concerning amount of the thinking in my use of the claude-4.5-sonnet model is dedicated to the Agent trying to solve this issue instead of solving my query. Needless to say, I’m very upset but it could be my fault.

Steps to Reproduce

Create an installation of the latest version of Ruby On Rails.
Tell the Agent “Execute this command in WSL ONLY.”

Expected Behavior

The Agent opens a console for wsl as per the Terminal: Default Profile set through the Command Palette and executes the command without the need of the “wsl” prefix in a powershell terminal

Operating System

Windows 10/11

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.0.34 (user setup)
VSCode Version: 1.99.3
Commit: 45fd70f3fe72037444ba35c9e51ce86a1977ac10
Date: 2025-10-29T06:51:29.202Z
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Windows_NT x64 10.0.26100

For AI issues: which model did you use?

claude-4.5-sonnet thinking

Does this stop you from using Cursor

Yes - Cursor is unusable

hi @John_McWhirter and thank you for the detailed bug report.

We have improved terminal for Agent and one of the changes is non-interactive/readonly terminal.

There are two main settings that may affect your usage.

  • Auto-Run Mode: did you use Ask Every Time, or Sandbox?
    (Sandbox separates Agent terminal)

  • Legacy Terminal - Allows terminal usage more closely resembling pre 1.7 functionality.

  • Additionally some reported that they had fixed access to some terminal commands through symlinks.

Please let me know which other recent and relevant bug reports you saw. It may help to identify and resolve issue faster.

Just echoing this problem and linking to a similar thread - this is still an issue on 2.0 and occurs regardless of auto-run mode or legacy terminal use.

This was an issue I saw <1.5, was resolved in 1.6.x, and then started happening again in 1.7.x and 2.0.x . Only reverting to latest 1.6 fixes the issue consistently.

1 Like

I am having this issue too, on all recent versions (2.0 now). The agent executes a command, i see the output while its running but when the command is finished the output dissapears, the agent doesn’t receive it.

This is extremely annoying because now i cant’t use any terminal commands.
I run mac Macbook pro M2 chip, on Tahoe 26.0.1 with the most recent Cursor version.

Please cursor team resolve this issue