Cursor v1.6 - Pre-release Discussion

Please resolve this issue quickly; I can’t see MCP’s response right now, which has significantly reduced my ability to manage conversations effectively.

@andrewh @condor
A very annoying bug: adding files from File Explorer to a new chat by clicking “Add the files to a new chat” closes the current chat and stops the agent.

Version: 1.6.3 (user setup)
VSCode Version: 1.99.3
Commit: 6ff61021bde05fe311c784a7d29c9320976128c0
Date: 2025-09-09T22:59:10.623Z
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 10.0.19045

When is Agent Windows used, and what exactly is the purpose of this feature? It seems like Cursor is developing features that nobody actually needs—features that don’t solve real problems while simultaneously causing essential functionalities to get lost.

Compared to Agent Windows, I find Copilot’s Open Chat in New Windows more practical.

Looks like you tried ’em all :laughing:

Hello together,

is it possible to get access to the pre-release versions of Cursor ai ?

Damian

press cmd e no longer opens up background agent if u have agent window open it opens agent window - it looks like a cleaner way to use cursor for agent mode with history in the agent sidebar and just cleaner I think this is the future look for gent mode that there aiming for

You can select Early Access there:
File > Preferences > Cursor Settings > Beta > Update Acess

@andrewh @condor
Every time Cursor launches, especially with the latest updates, it shuts down my whole computer and consumes all the resources. This is a desktop PC with a very powerful processor, and yet it still reaches 100%.

Hi,

Apologies for being mildly off-topic.. Whereabouts are you finding the download for v1.6 pre-release, please?

I’m only seeing up to 1.5.11 on the Cursor ‘latest’ track in the download API.

this is im assuming for their multi agent workflows. some of cursor’s eng talked about/hinted this

in ide, you go to settings, beta, earlyaccess

Agent Window seems a slightly different layout to make vibe coders happy? Definitely not needed.

Good thing you’re not forced to open it.

It’s actually a big step in the right direction and I hope to see that UI evolve even further.

in 1.6.5:

  • terminal commands take an extremely long time to execute (i’ll try models: gpt5, claude4sonnet, gemini2.5, grok-code-fast-1, auto-mode)

about 30 seconds for this (previously, the same thing was done in less than 1 second):

- 2 “grep” functions Search, what this?

@raythurnvoid I strongly disagree! :smiley:

I’ve been looking for a long time now for ways to break free from VSCode’s very lacking support for working with more than 1 monitor. I’ve always found it unbelievable that at this point VS Code, the de facto IDE everyone is using, has made so little progress with this (to be fair they’ve explained several times that there is little they can do VSCode being an Electron-based application).

Since the Agent chat panel has become a permanent part of the screen this has become even more acute, and its getting worse by the moment as I delegate more and more tasks to AI, now in the form of several local and background agents working —and potentially requiring my attention at any moment- simultaneously. It’s insane how many times per minute I have to toggle, open, close, or resize tabs, windows, settings, dialogs, etc.—and the cognitive load this adds -at least for me- to daily work is enormous!

This agent window seems to be an attempt at solving this, enabling a dual-monitor setup, with your human tasks on one side and all the Agent tasks on the other, each with enough space and without one UI constantly covering the other. Makes a lot of sense to me! :+1:

It’s already built in

File > duplicate workspace in a new window

Or

Ctrl p > split editor in a new window

Yes tried that already, but I never found a way of taking advantage of it, in fact it only increased my feeling of disorientation… I think I’d rather go with this clear -and automatic!- human/agents division… dont know, lets see what they come up with! glad to see they are investing in trying new concepts like this!

We’re looking at this - we see it too

It seems that when I do a cmd+f to find and replace, I can only search now and not replace.. no change if I go to Edit > Replace instead of cmd+f. I am running on macOS with a linux remote ssh env. Note that the arrow to the left of the find input isn’t there to expand into replace.