Cursor Grok 4.5 model does not make inline diff

Support Bug Reports
, ,
1

Describe the Bug

When using the Cursor Grok 4.5 model for code modifications, the inline diff is not displayed. Instead, the model applies the changes directly. I suspect this is because Cursor Grok 4.5 often creates and executes a separate Python script to modify the code, rather than editing the files directly, which appears to bypass the inline diff mechanism. The Composer 2.5 model, however, displays inline diffs as expected.

Steps to Reproduce

  1. Ask the Cursor Grok 4.5 model to modify some code and observe the result.
  2. Ask the Composer 2.5 model to modify the same code and observe the result.

Expected Behavior

  1. The code modified by the Cursor Grok 4.5 model does not show an inline diff.
  2. The code modified by the Composer 2.5 model does show an inline diff.

Operating System

Windows 10/11

Version Information

Version: 3.14.7 (user setup)
VS Code Extension API: 1.128.0
Commit: a758f2241ca99fecf380180b6cbdbbce0f1f42c0
Date: 2026-07-30T06:41:34.009Z
Layout: IDE
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.291
OS: Windows_NT x64 10.0.26200

For AI issues: which model did you use?

Cursor Grok 4.5

For AI issues: add Request ID with privacy disabled

98b842a4-8520-48e8-a47a-b73c5e8d495a

Additional Information

If Cursor Grok 4.5 does not use python tool for editing the code, I can see the inline diff. I think you have to stop using python tool for editing the code.

Does this stop you from using Cursor

No - Cursor works, but with this issue

4

Hey, thanks for the report. This is a known limitation, not something in your setup.

Inline diff and review only track changes the model makes via the native edit tools. When the model edits files through the terminal, like running a Python script, it writes straight to disk and skips the review layer, so the inline diff doesn’t show up. That’s why Composer 2.5, which uses the edit tools, shows a diff, but Grok 4.5 often doesn’t when it switches to scripts. The changes are still real and you can see them in Source Control and git diff.

What you can do right now:

  • Review those edits via Source Control or git diff, everything in the working tree shows up there no matter how it was changed.
  • Prompt the model more strongly to edit via the edit tools, not via shell scripts, for example add a rule that it shouldn’t write files via Python or other scripts.

Related thread with details: Agent edits made via the terminal (script/shell writes) don't appear in the inline edit-review UI

I’ll pass along that Grok 4.5 seems more likely to switch to scripts as a separate note about model behavior. If you can check the transcript to see why it chose to edit via a script, like tool calls failing or the model thinking it was cheaper, that’d help a lot. Let me know how it goes.