“I’m running Grok, but other models also don’t show the full reasoning process — part of it is greyed out. Is there a way to make it fully visible? That would be useful because you could interrupt the agent if it starts going in the wrong direction.”
1 Like
Hey, thanks for the question.
This is a known limitation. Reasoning models like Grok show a collapsed thinking process. A similar issue is discussed here: Thinking cycles cannot be expanded?
Try clicking the reasoning section (where it says “Thought for X seconds” or shows shaded text). It should expand. If that doesn’t work, or if it only shows minimal info, that’s a limitation of the current UI.
The team knows that full visibility into the reasoning process is important for developers. It’s critical for understanding what the model is doing and for tuning prompts.
Could you share:
- What Cursor version you’re on (Help > About)?
- Does the thinking block expand when you click it, or is it fully hidden?
- A screenshot of what you see would help.
I agree this would be nice to have. I will watch the thought process and catch something really insightful for a split second, but it will then be absent from the actual output.
@deanrie FWIW, there seems to have been some inconsistency in when thinking cycles can be expanded or not. Which was why I started my thread. I don’t know if it is just regressions, or if there is something else causing the issue. I’ve mostly been using the Claude models lately, which DO expand. That said, I noticed recently that the little arrow (chevron) next to the “Thought for Xs” no longer appears by default, and only appears if you hover or click it. These little ui indicators are important to helping your users navigate the ui, and the chevron (that right-pointing thin line arrow) should be visible at all times to indicate the Thinking block can be expanded to see the details. That used to be the case, but much like the Reasoning 
icon next to the model selections in the prompt editor, it seems to have disappeared in recent releases. I think adding these visual cues to the UI, while it seems small and potentially unimportant, is actually very important to assisting vision-based humans in navigating the ui quickly, efficiently, and consistently.
Thanks for the extra feedback about the chevron and visual indicators. That’s a helpful detail. The expand indicator really should always be visible for better navigation.
The team is already aware of the visibility of the reasoning process and UI indicators. I’ve passed along your feedback about how important these visual cues are.
1 Like