Cursor refusing to test plans

Bug Reports
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When i try to make a feature, i normally ask cursor to test the feature, making sure it works.
But lately, after 2.0, it refuses to test anything at all.

It just gives me a reply like this:
Not run (no direct access to the live dashboard in this environment).

Steps to Reproduce

I can simply say “Check this test.php for any issues, if any issues appear, fix, test, repeat until all issues are fixed”
Then i would simply get “edited, not run”.

Expected Behavior

I would expect it to run my commands.

Operating System

Windows 10/11

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.1.0-pre.27.patch.0 (system setup)
VSCode Version: 1.99.3
Commit: 482d448c4744a02d5d20a21f9b208cf0ad2b72e0
Date: 2025-11-13T08:17:39.792Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Windows_NT x64 10.0.26200

For AI issues: which model did you use?

Auto.

Does this stop you from using Cursor

Yes - Cursor is unusable

2

Hey, thanks for the report. This looks like a terminal tool configuration issue introduced in 2.0.

Please check these settings:

  • Cursor Settings → Cursor Settings → Agents → Auto-Run Mode - what’s selected?
  • If you’re using a custom mode, make sure the terminal tool is enabled

Try this workaround:

  • Settings → Chat → Inline Editing & Terminal → enable “Legacy Terminal Tool”

This helped in similar cases where the agent stopped executing commands after the 2.0 update.

Also, please share a screenshot with the full error in context. The “no direct access to the live dashboard” message is unusual and might point to something else.

Let me know if enabling the Legacy Terminal Tool helps.

3

Hello.

I’ve always had it set as “Run Everything”.

image
image957×327 13 KB

Also, always used the Legacy Terminal Tool
image
image952×451 16 KB

Still does not run.

4

Thanks for the extra details and screenshots. It looks like your settings are correct, which may point to a bug in the pre-release version you’re using.

Please help us debug this by sending:

  • A full screenshot showing the entire agent chat interface at the moment the error appears
  • Request ID: open the context menu (top right of the chat) → Copy Request ID for the failed request
  • Console errors: press Ctrl+Shift+I to open DevTools → Console tab → take a screenshot of any red errors that appear when the issue occurs
  • A check on the stable release: if possible, try the stable release (not the pre-release) to see if the issue still happens

This will help the team understand whether this is specific to the pre-release version or a broader terminal tool issue.

5

Sorry for the late response.
Been a while without issues.
But without changing my rules or anything, it came back.

Version: 2.2.0-pre.16.patch.0 (system setup)
VSCode Version: 1.105.1
Commit: 6384484c4763e240f454e603d06bfd60967dd500
Date: 2025-11-29T08:21:29.077Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Windows_NT x64 10.0.26200

It checked the location, etc.
Made a script, but it didn’t execute it.
It lets me execute it.
And even if i ask it to run the scripts, it just re-does the script or makes a new version.
Then tells me to run it.

CTRL + SHIFT + I = Open a new tab, not console.

image
image458×657 21.9 KB

image
image465×793 29.2 KB

59a04b6e-b5ae-4b8a-9d75-5c1139b64b0e

image
image466×841 8.51 KB

I even tried specifying “Run” in a rule now.
It still didn’t work.

image
image951×458 23.6 KB

6

Now it happen again!

image
image488×908 29.1 KB

image
image969×590 35.7 KB

7

Thanks for the update! I see the Agent is now successfully running commands - executing scripts, tests, and checking logs. That’s a good sign.

It seems the issue was temporary or tied to a specific chat. This could be:

  • A bug in the pre-release version that’s fixed in a newer build
  • A chat context problem resolved after a restart
  • A conflict with one of your Rules

To make sure everything stays stable:

  • Keep using the Agent in new chats and watch if the problem repeats
  • If you see “Failed” again, send the Request ID and a screenshot right away
  • Check if there’s a pattern - like it happens after a certain number of commands or context size

If the problem returns, I’ll escalate it to the engineers with the details. For now, it looks like a pre-release bug that resolves with a restart or update.

8

Forgot to reply here.
I made a new discussion post about it and other issues.

More and more being broken ever since 1.6 - Discussions - Cursor - Community Forum

9

It happened again.

image
image399×903 28.4 KB

Version: 2.2.0-pre.28.patch.0 (user setup)
VSCode Version: 1.105.1
Commit: 6371096c7c5fbb558bfc56e9e6438f4d1915fe70
Date: 2025-12-05T22:43:36.937Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Windows_NT x64 10.0.26200

Message:

Message

I’ll help you install RabbitMQ 3.13.7 with Erlang 26.2.5 on your AlmaLinux 9.6 server. Let me first check the current state of your system and then proceed with the installation.

[1 tool called]

Your server is running AlmaLinux 9.7 with no existing RabbitMQ/Erlang installations. Let me proceed with installing the specific versions you need.

[6 tools called]

Let me check the available Erlang versions and proceed with installation:

[1 tool called]

Let me try installing directly with the specific versions:

[1 tool called]

It looks like the commands are being cancelled. Would you like me to:

  1. Continue running the commands - I can proceed with the installation
  2. Provide you with a complete installation script - You can run it manually at your convenience

The installation process involves:

  1. Installing Erlang 26.2.5
  2. Installing RabbitMQ 3.13.7
  3. Enabling and starting the services
  4. Enabling the management plugin

Please let me know how you’d like to proceed!

Request ID:a7420184-ced7-4558-9e17-e87c7b44f031