Hi, I’ve noticed that the agent terminal is now read-only, and this has become a problem.
Many times, the agent successfully starts a server, but it doesn’t kill the process when needed. Since I can’t run commands directly in the agent’s terminal anymore, I have to manually find the process running on the local port and kill it with kill.
This workflow feels inefficient, and I don’t think it should be like this.
Why is the agent terminal read-only now? Is this an intentional change?
similar situation to me. The “sudo” command cannot be executed because it requires password.Many commands no longer display output, degrading the debug experience. I was compelled to consistently downgrade my cursor to 1.5.
this is still persistent for me too - any updates?
I really appreciate the ongoing development, but this change is a dealbreaker for my workflow.
I like to keep an agent terminal session open so I can interactively try commands, tweak setups, and debug in real time, especially when configuring environments or working through multi-step shell tasks. Losing write access to the terminal defeats the purpose for me.
Hey, @Mauroariza @Jay_Dholakia @chrisae9 @mengxinpy thanks everyone for the report. We’ll look into it. Also, could you let me know which OS you’re using and which terminal shells you use?
Cursor IDE
This is how i used to edit in the terminals, start expo, and debug. What is the workaround with the new update?
just using the new cursor update
MacOS
Version: 1.6.27 (Universal)
VSCode Version: 1.99.3
Commit: d750e54bba5cffada6d7b3d18e5688ba5e944ad0
Date: 2025-09-17T20:21:17.042Z (8 hrs ago)
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Darwin x64 24.0.0
Yes - Cursor is unusable
Version: 1.6.27 (Universal)
VSCode Version: 1.99.3
Commit: d750e54bba5cffada6d7b3d18e5688ba5e944ad0
Date: 2025-09-17T20:21:17.042Z (9 hrs ago)
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Darwin x64 24.0.0
I am not able to open expo as it requires typing a letter into the terminal to open the simulator..
Please bring back the edit capability in the agent terminals. We use it extensively to work with kubernetes clusters. We run few commands in the agent terminal to help the agent navigate the cluster, also provide temporary jwt authetication by running few commands manually in the terminal.
I’m not sure of macos, but can’t you just go to the Terminal tab at bottom and use trashbin to kill the terminal and process along it?
I’m on macOS 15.6.1 (Sequoia) and using zsh.
Thanks everyone for the info. I’ve already notified the team about this issue, and I believe it’ll be fixed.
In case it helps convince the product team to undo this change – one of my use cases is to run pytests of temporal workflows. Sometimes these workflows hang and I need to interrupt. The agent is able to pick up the interrupt/errors and fix them. However, with this change, I can ctrl-c but that gives the agent an exit status 0 so it thinks everything is working. This makes cursor completely unusable for me.
Is there an official place to download older versions? I know GitHub - oslook/cursor-ai-downloads: All Cursor AI's official download links for both the latest and older versions, making it easy for you to update, downgrade, and choose any version. 🚀 exists but all the ads make me wary of using it.
Version: 1.6.27 (user setup)
VSCode Version: 1.99.3
Commit: d750e54bba5cffada6d7b3d18e5688ba5e944ad0
Date: 2025-09-17T20:21:17.042Z
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Windows_NT x64 10.0.26100
I can’t even copy from terminal window anymore. Ctrl-C is now interpreted as an interrupt command instead of copy command. I used to be able to highlight and copy lines (error messages mosstly) and pste to chat. Previously only when the cursor is on a new line Ctrl-C is an interrupt. Now Ctrl-C is an interupt everywhere
Now Ctrl-C is an interupt no matter what my intended usage is …
Don’t get me wrong the the terminal workflow is now 10x better & faster as of this update
My turnaround times on dev and testing are vastly improved thanks to the terminal improvments
Same here, our entire R&D can’t work with cursor 
Can you please provide an ETA for the rollback/fix? It’s really a deal breaker in our case
It has completely disrupted my workflow… Why are you trying to block interaction with a terminal? It makes no sense! I am a DevOps engineer and I need to interact with a terminal 80% of the time (killing processes, entering values, etc.). Please rollback this change.
My workflow has been severely disrupted by this update. I’m unable to re-run commands, make edits to Agent-generated commands, and work alongside the Agent. The terminal is to be run by me, not the Agent; the Agent is here to support me. I shouldn’t be blocked from my own terminal.
This is not ideal.
In case you guys want to rollback while the fix isn’t available, I found this github repository with legit links from download.cursor.com
Update: Version 1.6.6 seems to be the latest without read only agent terminals
Same here, this is quite a significant and unfortunate change. Many people rely on terminal commands and need the flexibility to adjust them.
Yeah, I totally follow you. This is a regression. Loosing control of the terminals that can be interactive really really ■■■■■. Why oh why would Cursor do this?!?