Missing “Open in Terminal Pane” for Agent command output in newer Cursor versions

[Ali_J_khan]

I’m seeing a workflow regression in newer Cursor versions and wanted to check if others are experiencing the same thing.

In Cursor 3.0, when the Agent ran a terminal command, the command card menu had an option like “Open in Terminal Pane.” That was very useful because I could let the Agent run commands in chat, but when the output became long or important, I could move/open it in the normal bottom terminal pane and inspect it properly.

In newer versions, around 3.6/3.7, I no longer see that option. The menu now mostly shows things like auto-run, sandbox, allowlist, and copy command. The output stays trapped inside the chat command card unless I manually copy the command and re-run it myself.

This is painful for a few reasons:

1. Long logs are much easier to inspect in the terminal pane than inside a chat card.

2. It breaks the normal debugging loop: Agent edits code → Agent runs command → I inspect terminal output → continue.

3. It is harder to tell what actually ran, whether the command is still running, and what the full output/error was.

4. Copying and re-running commands manually defeats the point of Agent execution.

5. For people learning/debugging code, seeing the command output in the real terminal pane is much clearer.

Is this option intentionally removed, renamed, or hidden behind a setting? I tried looking under Settings → Agents → Inline Editing & Terminal, including Legacy Terminal Tool, but I still don’t see the same clean workflow from 3.0.

Expected behavior: Agent command cards should have an “Open in Terminal Pane” / “Move to Terminal Pane” option, especially for long-running or long-output commands.

Actual behavior: output stays in the chat command card, and the only reliable option is copying the command and manually re-running it in the terminal.

Would appreciate clarification from Cursor team or anyone who found a setting/workaround.

[Colin]

Hi @Ali_J_khan!

Thanks for the feedback! It does seem to have disappeared (I liked it too).

We’ve had some internal feedback about this recently, and there’s an open ticket on it. In the meantime, it is possible to send the running terminal to the background.

You can then click on the backgrounded terminal to view its output in the terminal pane.

If there’s any movement on bringing back “Open in Terminal Pane”, I’ll let you know!

[Ali_J_khan]

Thanks, Colin. That helps and confirms I wasn’t just missing a setting.

Now, I tested the background terminal workaround, and I think there are still two gaps compared with the old “Open in Terminal Pane” workflow.

First, the workaround seems to apply mainly to long-running commands. When the Agent runs a long-running task, I can see it as a background terminal, click it, and view output in the terminal pane. But once the task completes, that background terminal seems to disappear. In older versions, I could still see the completed Cursor/Agent command history in the terminal pane, in order, which was useful for reviewing what ran and debugging after the fact.

Second, for short/immediate commands that return output quickly, I don’t see a way to move/open that command output in the terminal pane at all. The output stays in the chat command card, and the only workaround is copying the command and manually re-running it.

That creates a real workflow issue:

1. Long outputs are harder to inspect inside chat cards. and disappear before i can read anything and i can’t see a logical history of what run. this seems totally contradictory to an agent session and inspecting their command flow (if I have to open different panes for each command, which even that i can’t seem to do be able to do).

2. Completed command history disappears or is not easily reviewable.

3. Immediate command output has no terminal-pane path.

4. The old workflow let me keep a clear, ordered record of Agent commands/output in the terminal pane.

5. Copying and re-running commands manually breaks the Agent debugging loop.

So my question is: is the team actively planning to restore the “Open in Terminal Pane” / “Move to Terminal Pane” for both running and completed Agent commands, including short commands? Or is the newer behavior intentional, which would be a really bad idea!

I would strongly argue this should come back. For debugging and learning, being able to inspect Agent command output in the actual terminal pane is much clearer than having everything trapped in chat cards or having the terminal appear briefly and then go away!

Thanks again for confirming there is an open ticket — please add this specific distinction to it: background terminal helps with active long-running commands which disappear once the task completes, and it does not solve the issue of viewing completed command history and/or show immediate-output commands.

[Ali_J_khan]

@Colin. Any updates/thoughts?

[Colin]

First of all, I agree with your reasoning!

Seocndly, no news since last week!

We have to choose our battles, and bringing this functionality back isn’t a high priority right now. If that changes, or if there’s movement on the ticket, happy to update this thread.

[Ali_J_khan]

tell you what, I appreciate the response man. Ty. I hope my team and I don’t have to leave Cursor because of this. The reason we picked Cursor over Lovable or and other junk is because of it’s a natural extension of VS code and how it catered to making devs more efficient as opposed to vibe coders making slop.

I hope Cursor doesn’t forgo what we felt was its core demographic.

[Niten_M]

I’d like to add my support for bringing back the “Open in Terminal Pane” option to all running CLI commands. It feels like the guiding principle with this move (also hiding “Thoughts”) has been to “abstract as much of the messy code stuff” from the user as possible.

We are not ChatGPT users. This is not Microsoft. We are systems & software developers who need to maintain visibility into what the agents are doing regardless of how good the coding agents become.

It’s the equivalent of having a junior employee refuse to show you their work b/c “trust me broh”.

I chose Cursor originally b/c it wasn’t promoting the whole “vibe coding” ethos & actually helped me create standard developer documentation like TDD & FDD which the agent would happily use to produce well designed & informed code. I don’t want an “auto-majik” tool that hides the complexity of software development from me. Leave that to Microsoft. I need to be fully aware of the agent outputs, or I simply can’t use it for system tasks, networking, or deployment tasks which can’t simply be undone by “reverting git”.

This should be a high priority. We need an open & transparent coding assistant to help us produce higher quality code efficiently. Not a “click & pray”, “OpenClaw style”, “Lights out” robot to code everything & leave us out of the loop.

[Tom_Coustols]

hey @Niten_M, I don’t know if this really answers your worry, but you actually have the ability (in latest version) to modify the tool details. I know this isn’t all of it, and it would be great to at least have some words about it, but until then, know that this exist! (see gif)

[Ali_J_khan]

This really is a case study in how VC money and pressure makes companies shoot themselves in the foot. Abtracting away the thoughts stream and terminal pane is such a sus move. And this coupled with those weird cursor sponsored ads of people making meaningless apps like ‘i had an idea to make a an app that changes the icons on my laptop’ ok but why? and good luck scaling that stuff.

The entire point with cursor was to make developers more efficient. not make them become subservient to AI. And Cursor already had the ‘agent view’ for end-to-end ai driven coding for the non-tech demo. Their changes in the editor area have been at best incremental and if anything counterproductive in light the lack of terminal view and thought streams. My team and I have been looking at claude code cli and cursor’s cli as better alernatives, though we still hope that Cursor takes a step toward transparency again. Abstraction should be an optional choice, not an enforced philosophy, at least for devs. AKA, Cursor should go ‘you can use settings to remove thoughts, terminal view, etc. but you don’t have to!’

[Abhiram_Thrivarna_Si (Abhiram_Thrivarna Siddoju)]

Bringing back “Open in Terminal Pane” should be a high priority for the Agent terminal experience. In Cursor 3.0 this was one of the most useful parts of agent workflows: I could let the agent run commands in chat, then move long or important output into the normal bottom terminal for proper scrolling, searching, copying, and post-run debugging. Since ~3.6/3.7 that option disappeared from the command card menu, and the current menu (auto-run / sandbox / allowlist / copy) does not replace it. Backgrounding only helps for long-running jobs, and even then completed agent terminals often disappear, so short commands and finished sessions still trap output in the chat card and force a copy/re-run workaround. Please restore “Open in Terminal Pane” (or “Move to Terminal Pane”) for both running and completed agent commands — this is core power-user UX, not a nice-to-have, and losing it has clearly degraded day-to-day productivity for people who rely on agent-driven terminal work.