Add a setting to keep Request & Response permanently expanded in Agent chat

Ideas Feature Requests
,
1

Feature request for product/service

AI Models

Describe the request

What I’m Requesting
Please add one or more of the following options:

  1. A toggle in Settings → Agent / Chat

“Always expand requests” (on/off)
“Always expand responses”
“Remember expansion state for each message”

  1. A right‑click UI option on the Request/Response panels

“Keep this expanded”
“Expand all” / “Collapse all”

  1. A keyboard shortcut
    Example:

Expand all agent details: Cmd+Shift+J
Collapse all: Cmd+Shift+0

(These exist for folding code in the editor, but not for the Agent panels.) [forum.cursor.com]
4. A developer mode
A toggle like:

Developer: Show full request/response JSON expanded by default

This would be extremely valuable for MCP developers debugging their tools.

Why This Feature Matters
:check_mark: Speeds up debugging
Agent requests/responses often contain long JSON structures, tool parameters, and reasoning output. Having them collapsed slows inspection.
:check_mark: Critical for MCP development
MCP tool authors rely on request/response visibility. Current UI hides critical information.
:check_mark: Improves transparency
Many users want to see exactly what the model sees without clicking 10 times per message.
:check_mark: Aligns with Cursor’s philosophy of developer‑friendly tooling
Cursor proudly supports MCP, deep context, and transparent debugging — this feature aligns perfectly.

Supporting Evidence

Many users rely on MCP logs because the UI collapses request/response details.
Cursor stores full MCP logs here:
~/Library/Application Support/Cursor/logs/…/Cursor MCP.log
This log shows raw I/O that the UI hides.

There is no existing setting in Cursor to control expansion/collapse behavior.
No documentation or feature mentions persistent expansion; users rely on logs instead. [forum.cursor.com]

Cursor already exposes hotkeys for collapse/expand in the editor, so a similar UX pattern can be applied to Agent chat.

Conclusion
A simple UI/setting update would dramatically improve developer experience for:

MCP developers
AI tool creators
Users debugging complex responses
Anyone who wants full visibility of model I/O

Please add a permanent “expanded mode” option for Request & Response panels in Cursor’s Agent chat.
Thank you!

3

Why stream of thought matters to the Human in the loop.

This feature request is important to agent orchestration, as well working across monorepo and multi repo workspaces.

I work with Claude mostly for Work, but prefer Cursor for my personal side projects. Both Claude (IDE extension) and Cursor chat interface have their own pain points and pro/cons in developer & workflow interface experiences.

I like to read the stream of “thought” the “reasoning” and often I have found it helps to catch things that I may have been unclear about, or that the best-guess scenareo of the LLM Model jumped to the incorrect conclusion / weighted path.

I find being able to read the agent’s intermediary feed gives me insights into what I missed and I can reframe my prompt or fix my ADHD sentence structure and directives or Tasks to more precision.

When the cursor chat is "reasoning” it streams by in a 200-300px height view in lite text and I can catch glimpses of it to process cognitively myself, but it’s too quick and as soon as the agent is done reasoning, it replaces the stream window with its output composition stream.

I NEED a way to have cursor let me keep those streams of reasoning visible and not get replaced. Please consider his feature. Happy to discuss my workflow over a chat if your Graphical HC Interface Experience team would like to research. :slight_smile: