Response details for Agent replies in Cursor 3.0

Feature request for product/service

Cursor IDE

Describe the request

I would like to suggest a small but very useful improvement for Cursor 3.0 Agent replies.

It would be great to have extra details shown for each agent response, such as:

  • Copy as Markdown — copy the full raw markdown of the reply, not only the rendered text
  • Token usage — show how many tokens were used for the request and for the response
  • Execution time — show how long the request took from start to finish

Why this matters

In Cursor 3.0, the workflow is much more focused on agents, review, and iteration. Because of that, an agent reply is not just chat text anymore. It is part of the working process.

Having these details would make replies much more useful.

1. Copy as Markdown

This would help when reusing agent output in:

  • documentation
  • GitHub issues
  • pull requests
  • notes
  • task descriptions
  • external tools

Right now, it is often much more useful to copy the original markdown instead of copied rendered text.

2. Token usage

It would be very helpful to see:

  • input tokens
  • output tokens
  • maybe even total tokens

This would help users better understand:

  • prompt efficiency
  • context size impact
  • model usage
  • cost awareness

This is especially useful for people who actively compare workflows and try to reduce unnecessary token usage.

3. Execution time

It would also be useful to see how long the reply took.

For example:

  • was it a fast answer
  • did it take longer because of tools
  • did large context slow it down
  • was the agent doing more reasoning steps

This gives better visibility into how the workflow behaves.

Suggested UI

This could be shown in a small row under each agent response, for example:

  • Copy Markdown
  • Tokens: 2.8k in / 1.1k out
  • Time: 14.2s

I think this would be a very strong quality-of-life improvement for Cursor 3.0, especially for users who work with Agent mode every day.

Screenshot / Screen Recording

Operating System (if it applies)

Windows 10/11
MacOS
Linux

It would also be great to bring back context loading visibility, like it existed before.

For example, showing how much of the available context is currently used, in percentage form.

This was very useful because it helped understand how close the current chat or agent workflow was to the context limit. It made it easier to adjust prompts, reduce unnecessary context, and better predict when performance or response quality might drop.

I think this would fit very well into Cursor 3.0 together with token usage and response metadata.