Queue follow-ups after “thinking” ends, not only after the full reply

[brighty_dev]

Feature request for product/service

Cursor IDE

Describe the request

Summary:

Today, queued messages run after the current agent turn finishes (full answer, tools, etc.). With thinking models, you often see enough in the Thought / reasoning block to steer or correct the next step early, but you still wait for the rest of the stream and work to complete.

Request:

Support an option (or a queue mode) so a queued user message can be applied as soon as the thinking phase for the current turn completes, instead of waiting until the entire reply and agent loop for that prompt is done. That gives a faster steering window without replacing normal “queue after full completion” behavior for users who want it.

Why:

Shorter feedback loops, fewer wasted tokens and edits when reasoning already shows a wrong direction, same queue UX but with a finer checkpoint than “turn done.”

Operating System (if it applies)

Windows 10/11
MacOS
Linux

[kevinn (Kevin Neilson)]

Hi @brighty_dev Thanks for the request! I think we actually already have the feature you’re looking for. When the agent is working on something and you send a new message, it is by default queued, yes, until after the current model finishes thinking and completes the task. However, you can “force” the queued message to be send immediately by pressing send now - and the agent will immediately consider the new message, interrupting its current thinking process.

Let me know if you have any follow up questions!

[brighty_dev]

Thanks Kevin, yes I know the force steering and it is useful, but I was thinking more of non-forceful steer so that it can pick up after first reasoning/ thinking on its own.