"Cmd+K Has Been Retired" from terminal?

I was shocked seeing the “Has Been Retired”. Honestly this was a powerful feature, standing out against competitors and make it a complete AI integrated IDE. Was waiting for you guys to bring a standalone terminal version like this. I would only use that terminal!

This feels so wrong, please bring it back @Cursor

Probably leaving Cursor if this feature isn’t brought back to live…

Please bring cmd+K back. The agent is very clunky when all you need is directed in-line edits. This is the reason my company uses Cursor over competitors. We are trying out competitors now.

Seriously… what the actual heck, why was this extremely useful feature removed? One of my favorite and most used Cursor features.

No, having the agent do it instead is NOT a good replacement. It’s way more clunky, takes longer, and consumes WAY more tokens. Really hard not to suspect that increasing token use (and thus $$$ for Cursor) wasn’t the motivation here…

Edit: Wow, so useful :roll_eyes: what would have taken 1 second in the console wasted minutes with me watching the agent spin its wheels before I stopped it and just ran the darn command myself.

This is a bummer.

Especially in the terminal, cmd+k had been a huge productivity boost. There are a lot of CLI commands where I can’t remember (or don’t want to type) the exact parameters. But it was super easy to say, “push and set upstream” and have Cursor figure out the details. And being able to see the pasted command before running it gave me the confidence that the command was correct (and let me correct it easily when it wasn’t).

Moving to another window which then takes longer and requires copying and pasting the command (I don’t want it to run without confirming it’s what I want), is not nearly as simple a flow. Yeah, it works, but I use Cursor because the UX is seamless, not just because “it works.” By far, Cursor’s main benefit for me has been the thought put into the UX, and how it made my workflows easy. Changes like this are chipping away at the experience and leading me to consider what other alternatives are out there.

Small things can be huge, when it comes to UX. Please reconsider retiring cmd+k at least from the Terminal.

Using Cursor for the last 2 years. Cmd+K is the reason I have stuck with Cursor. Please restore it.

Same thing here, I was using this features multiples times a day. The CMD+L is so much slower. Please bring this feature back because it would be enough for me to leave Cursor as my software of choice..

Please restore this feature, the most important reason to use cursor!

I created an account on the Forums just to reply to this.

This is a brutal feature cut. ctrl+l does not replace ctrl+k, they serve different functionality.

Please, talk to your manager / product owner to reconsider. If they integrate the agentic panel into the ctrl+k panel to reduce implementation differences, fine, but a blazing fast + dumb + easy-to-use prompt window offers something a thoughtful, smart, time consuming to use window cannot

I also created an account to comment on this. Cmd-k was an essential feature and it should be restored. If not, I am going to switch to a competitor.

@Colin do you know if this will be restored? It’s significantly slowing down development speed for the agent to make small narrowly-scoped changes (e.g., highlighting a few fields and prompting “convert this into a struct”)

Please put cmd+K logic back, I mostly stayed on Cursor, because that feature where I can quickly ask about some part of the code or do quick refactoring of the small part.

I was really disappointed when 've seen “Cmd+K Has Been Retired”.

This has got to make a comeback, interrupting agent flows or spinning agents just for ad hoc commands is nonsense. Please bring it back asap

Hi all, the “why”: Use of Cmd+K had dropped a long way from its peak as more of this kind of work moved into Agent, and keeping a second inline system maintained alongside Agent was no longer something we could justify, so it was folded into Agent.

A lot of the extra wait comes from the model. Most Agent setups run reasoning models that think before they answer. For quick edits and shell commands, pick a lighter option in the model picker, for example Composer 2.5 with Fast on, Grok 4.7 with effort on Low and Fast on.

If you’d like to see a command before it runs, set Cursor Settings > Agents > Execution and Approvals to Allowlist with nothing in the list, and Agent should ask before running anything.

I’ve passed along the feedback from this thread to the team.