New version hides agent tool call details in defiance of setting

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

In the most recent version, 3.10.20, agent tool call information is buried even further in the agent window with an ‘Explored (x) tools’ message (where ‘x’ is the number of tools explored/used), making it even more difficult for users to see the agent and MCP interactions that are happening.

This is in defiance of having ‘Conversation Density’ set to ‘Detailed’, which should fully expand these interactions in the agent window.

Steps to Reproduce

  1. Update to 3.10.20
  2. Set Conversation Density (“Choose how much detail agent tool calls show in the conversation”) to Detailed.
  3. Try to make a tool / MCP call.
  4. Observe that the setting isn’t honored, and all information is now obscured underneath an ‘Explored…’ toggle that must be expanded several times before agent / MCP interactions can be seen.

Expected Behavior

All tool calls and MCP interactions should be fully expanded when the ‘Detailed’ setting is set. However,

Screenshots / Screen Recordings

Screenshot from 2026-07-09 06-44-18.png

Operating System

Linux

Version Information

Version: 3.10.20
VS Code Extension API: 1.125.0
Commit: 23b9fb205fe595ea2be29da7214e19762d037fc0
Date: 2026-07-07T07:03:33.071Z
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.256
OS: Linux x64 6.8.0-124-generic

For AI issues: which model did you use?

N/A

Additional Information

Meta comment: I’ve filed a variant of this bug on at least one other occasion. Beyond this bug, it appears that Cursor has made a decision to actively abstract and hide useful info from users – and this seems to be getting worse in recent versions.

Especially for users in Editor mode (who are presumably experienced developers), the default should be to show everything – all command line output, all agent output by default, rather than dumbing down the interface to appeal to less technically inclined users who’d likely use the Agent window anyway.

Does this stop you from using Cursor

No - Cursor works, but with this issue

Hey there!
The Detailed setting isn’t being ignored. It’s working, but it controls a narrower scope than you’re expecting.

Here’s what “Detailed” actually does today:

  • It fully un-groups file edits and terminal commands so those show inline.
  • Read, search, and exploration steps (file reads, grep, list, glob, web fetch/search, and Cursor’s internal “list available MCP tools” lookup) always fold into the “Explored N tools” summary at every density, including Detailed. Click it to expand. That’s the current design, not the setting failing. The changelog even describes Detailed as “near-complete step-by-step context” rather than fully expanded.
  • Your actual MCP tool calls render as their own rows and aren’t hidden under “Explored.” If a specific MCP call is getting swallowed into that summary, tell me which one and I’ll dig in.

On the bigger picture: a real “expand everything by default” mode for Editor power users is a fair ask, and it’s the same thing you raised back in Uncompacting agent command execution. There’s an open request for exactly this you can add your voice to: Add a setting to keep Request & Response permanently expanded in Agent chat. I’ll capture your specific case (exploration + MCP-list calls staying collapsed in Detailed) as a feature request too, and we track threads like this to gauge interest and help prioritize.

Hi @mohitjain, That’s helpful context! I added my vote to that open request, and hope the team will consider acting on it. Having a human in the loop to supervise agent actions is critical, and that includes being able to see what the agent is reading, searching and exploring, so I respectfully disagree with this new default setting, especially for the power-user-friendly Editor mode.

Regarding our MCP, I have observed MCP tool calls being hidden; I’ll capture a screen video later today and include it here.

Not to be overly dramatic, but I’ve been working all morning with this change, and it’s driving me up the wall. I really feel like I have less control over the agent and its work. As much as I love Cursor, this is honestly the first time that I’ve considered looking at other AI assistant tools that don’t hide all the details.

Please reconsider this change, and prioritize adding the ‘show all’ setting.

i feel the same…i was a really satisfied cursor-fanboy, but taken away the ability to see whats going on makes me mad. telling cursor with hooks and rules that i ALWAYS, unconditionally, want to see with tail -f what is happening leads to…whatever, sometimes doubled terminal sessions which savbotage each other, or no output at all. this is how you render a beautiful product unusable. ok, maybe exaggeration, but not seeing what goes wrong…no option.

This is critical.

Even MCP tools custom UI made precisely for the user to see them when those tools are called (MCP Apps protocol extension) get collapsed and hidden, even when conversation density is set to “Detailed”. This is a huge bug.

@trevorSmart on your MCP Apps case specifically: what you’re showing isn’t intended behavior. Your ui:// widget renders fine while the turn is live, but once the turn completes it gets rolled into the “Worked for Ns” work-group and the widget disappears with it, even in Detailed density. Density controls a different layer (edits + terminal commands); it doesn’t govern this work-group collapse today. We’ve let the team know, and this is an issue we’re tracking. No timeline to share yet, but I’ll post here when there’s an update. Interim workaround is just to expand the “Worked for Ns” header when you want the widget back.

@Andreas_Wittemann sorry it’s been getting in your way. Your core point - that Detailed today doesn’t expose exploration and MCP activity - is the same one @rsb is pushing on, and it’s on our radar, we track threads like this to help prioritize. The tail -f / doubled-terminal side-note sounds like a separate issue from the collapse behavior; if you can spin that up as its own report with a small repro (which hooks/rules, what commands, what you see) it’ll be much easier to look at.

Another annoying mandatory new feature. I wish I could roll back to an old version. You keep making the UX more difficult with every release. Please do not touch features unless you provide an option in the config. Now, I have to spend so much extra time to find the AI slop in the chat window.

@Adnan_Boz - there is a setting for this today, though it doesn’t go as far as you’re asking: Cursor Settings → General → Layout → Conversation Density, set to Detailed. That pulls file edits and terminal commands out to the top level so they show inline. What it doesn’t do yet is expand the reads, searches, and MCP calls that fold into the “Explored N tools” summary, which is the gap this thread is about. We’ve let the team know and it’s an issue we’re tracking.

@rsb one correction to what I told you earlier: you were right. MCP tool calls do get folded into the “Explored” group alongside reads and searches, at every density including Detailed. I said otherwise back in my first reply and that was wrong. So no need to spend time on that recording unless you’re seeing something beyond this.

It is already set to Detailed. That is not the feature I am referring to.

The problem is that the “Worked for…” sections started collapsing after I responded in the chat. I think this began a few weeks ago; it worked fine before.

In a long chat, it is very time-consuming to find the “Worked for…” sections and expand each one to see what changed.

One of Cursor’s best features is how it displays diffs and agent changes. Without this feature, Cursor feels no different from Codex to me.

@Adnan_Boz Just ran into this issue a moment ago, and it really threw me off. Hundreds of lines of important thinking and output compressed into ‘Working for 1m 15s’ which makes a scan of the agent chat incomprehensible (question to code updates separated by this single line which eviscerates the solving process in a casual read).

@mohitjain et. al, these may be technically separate issues, but you’re hearing a plea from some of your most dedicated users in this thread to stop hiding information that was shown by default mere weeks ago. It doesn’t make Cursor better. Please nip this unfortunate set of design choices in the bud.

Hey @trevorSmart,

A fix for MCP App widgets disappearing into the collapsed “Worked for” section has landed and will be included in an upcoming version of Cursor (3.27). With it, the widget stays visible in the chat after the turn finishes, and also when you switch away and reopen the chat.

Once you’re on that version, give it another try and let me know if a widget still gets folded away.

@rsb @Adnan_Boz this change is specific to MCP App widgets, so it doesn’t change how other tool calls fold under “Explored” or “Worked for”. Your feedback on that is still noted here.

I confirm the bug is gone, thank you @Colin