Hey everyone!
v1.6 is now available to all users! 
The full changelog can, as always, be found here:
Feel free to use this thread to share your feedback, or create a new post in Bug Reports for support with any bugs or issues you may find.
I have been working with many minutes with it already. Thus far all good, thanks! 
No really, I think it improved rule following, mcp tool setting is now showing my database tables (not sure why) and this far agent has not stuck yet.
Can you check accessibility settings?
Did this update break –user-data-dir={path\to\folder} support for cmd line launcher on windows? The default user-data dir writes context to my HD instead of my SSD which makes cursor agent absolutely unusable after about 10 messages
I’m sorry, I don’t fully understand.
Could you point me exactly to where and what I should check?
note @danperks, whilst the changelog now reflects 1.6, the downloads page isn’t yet up to date: Downloads | Cursor - The AI Code Editor
Drag and drop from windows explorer still isn’t working 
more details below
Very dissatisfied… I can’t use many extensions now. This is very, very bad. VS Code is only truly useful because of its extensions; without them, it’s just a piece of junk.
Please fix the behavior of GPT 5 models. They generally program beautifully in Cursor. However, if, for example, I deploy something and the result ultimately fails, the model sees the result as successful and draws a completely incorrect conclusion about what’s happening.
Also, it often happens that the model claims it made some changes to the project, when in fact it simply opened a couple of files, looked through their codebase, and made no changes—again, reporting successful changes.
Could you share any further detail? What isn’t working?
‘Keep all’ and ‘Review next file’ buttons unresponsive after the updates…
Looking into this. Could you try restarting Cursor to temporarily fix?
Issue 1: Terminal shows Success even when commands fail
Steps to reproduce:
-
Run a build, deploy, or any other command in the Cursor terminal using the model.
-
Ensure the command fails with an error.
Expected result:
The terminal should display the error and the model should recognize that the process failed.
Actual result:
The terminal shows Success, and the model assumes the process completed successfully even though it didn’t.
Issue 2: Changes are not saved before build/deploy
Steps to reproduce:
-
Enable autonomous terminal execution with AI (Run Everything).
-
Let the model modify the project codebase.
-
Allow it to trigger build/deploy.
Expected result:
The model should save all code changes before running build/deploy commands.
Actual result:
The model doesn’t save changes, so build/deploy runs on outdated code.
Issue 3: Terminal hangs after successful completion
Steps to reproduce:
-
Run build/deploy in the Cursor terminal.
-
Wait until the process finishes successfully.
Expected result:
The terminal should exit cleanly and show completion.
Actual result:
The terminal stays in a “loading” state indefinitely until I manually press Enter or Shift + Cancel.
Issue 4: Agent stops prematurely if terminal is closed
Steps to reproduce:
-
Run a process in the Cursor terminal (not necessarily build/deploy, e.g., uploading into the project).
-
Let it finish successfully.
-
Close the terminal using Shift + Cancel.
-
Wait for the agent to continue applying TODO changes.
Expected result:
The agent should continue making all agreed-upon TODO changes after the process completes.
Actual result:
The agent says “everything is done” even though not all TODO changes are applied. I have to remind it to continue.
I don’t know, what the advantage of Agent window. it is because my usage is wrong? As I see, it is just only the layout different to normal window.
So I want to staff show me the typical usage of Agent window
thanks. 
I just upgraded to 1.6.26 for MacOS, and now I can’t use it because I rely heavily on running terminals, manipulating commands within its sessions, and incorporating some chat context with echo. The new feature, “Agent terminals are read-only,” has made the cursor unusable for me (ssh and local).
Instead of implementing “Agent terminals are read-only,” I suggest flagging and halting dangerous commands to notify users about their potential impact, allowing them to decide whether to proceed at their own risk.
Understood - are you able to add any prompts/context/rules that could help? We have a feature called hooks coming that will also let you have better control over terminal comamnds
For some reason, it works by clearing all rules, memories and closing all projects. Uninstall and reinstall back 1.6.26. Now i can edit session terminal of agent and no longer shows “Agent terminals are read-only”.
My chats aren’t working after I reopen Cursor after closing my laptop. When I send the message, nothing is showing up. The stop button does pop up though. Restarting Cursor doesn’t help. I’m on the latest version of Cursor rn. Any chance this is due to the new update?
EDIT: Ah it might be my SSH upload.download speeds.
Version: 1.6.26 (Universal)
VSCode Version: 1.99.3
Commit: 6af2d906e8ca91654dd7c4224a73ef17900ad730
Date: 2025-09-16T17:12:31.697Z (12 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 arm64 24.6.0
I’ve mentioned this several times, and this version still hasn’t fixed it. I never get this error with the API provider when using ‘cline’ or ‘ro-code’, but it happens constantly with Cursor. This is a very bad experience.
Please fix this.
Think about it with me!! Wouldn’t it be better for us to use our own API key instead of your inference service? That way, the user doesn’t generate costs for you, right? So why not implement an automatic retry mode—at least 10 times—like ‘cline’ and ‘roo’ have, for example?"
Request ID: 8e17cc7c-db01-49fd-9cf3-793382e0c8c0
{“error”:“ERROR_OPENAI”,“details”:{“title”:“Unable to reach the model provider”,“detail”:“We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.”,“additionalInfo”:{},“buttons”:,“planChoices”:},“isExpected”:false}
ConnectError: [unavailable] Error
at wyl.$endAiConnectTransportReportError (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:4910:339807)
at Zvr._doInvokeHandler (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:487:36019)
at Zvr._invokeHandler (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:487:35761)
at Zvr._receiveRequest (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:487:34526)
at Zvr._receiveOneMessage (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:487:33348)
at WCt.value (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:487:31442)
at ye._deliver (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:2962)
at ye.fire (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:3283)
at Vot.fire (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:4895:12154)
at MessagePort. (vscode-file://vscode-app/c:/Users/decsters/AppData/Local/Programs/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7028:18292)