Cursor 3.1 is here! It’s been 10 days since we launched the Agents Window in Cursor 3, and this release brings a round of quality-of-life improvements based on how you’ve been using it.
Tiled layout: Split the Agents Window into panes to run and manage several agents in parallel. Drag agents into tiles, expand to focus, and use keybindings to navigate. Your layout persists across sessions.
Upgraded voice input: Dictation is now more reliable with batch speech-to-text. Press and hold Ctrl+M to speak, with a new recording UI showing a waveform, timer, and cancel/confirm controls.
Diff to File Selection: Jump from the diff straight to the exact line in the file. Once there, you have the full power of the editor, make changes manually, use Tab, go to definition, and more.
Plus many more improvements and fixes in the full changelog.
We’d love your feedback!
How are you liking the Agents Window?
What do we still need to refine?
How is the tiled layout fitting into how you work with agents?
If you’ve found a bug, please post it in Bug Reports instead, so we can track and address it properly, but also feel free to drop a link to it in this thread for visibility.
Thank you for the updates! I find the tile layout and upgraded voice input especially useful.
I’m curious – how many developers do you have on the team working on Cursor, and how extensively do you use AI to ship features and fix bugs? Your pace of development is staggering. When I compare your speed to your competitors, I am really interested to know: how do you do it?!
The Agents Window seems like it would be very good for my workflow of working with files and source control on one screen and agents on another. Since I moved to Cursor, I had been struggling with several issues to try to get that functionality:
While trying to find workarounds to those issues, I ended up returning to Visual Studio code for viewing files, diffs, manual edits, and source control, while having Cursor open in another window just for the agents. Those issues listed above have since been resolved, but there are two more unresolved issues that make Visual Studio Code superior for viewing diffs and managing source control, at least for my workflow:
Once those are resolved, then I can fully switch to Cursor and use the Agents Window as intended.
Its getting better! Initially I was not able to use it at all and now I spent whole day using it.
Better right panel, it tries to fit:
Git / PR view is best when narrow
Web browser is best when wide
List of changed files is best when narrow
Diff for a file is best when wide
Terminal is good either way, but can we get a separate appearance for it please so we can have dark terminal in light theme, as in real life
If I to suggest - restrict right pane to what looks best when narrow and give option to tab or stack vertically
Please keep improving git / PR view it gets better! Add more ci info, checks and deployments from github
Would be nice to be able continue working on same task but open new chat. For example open new chat tab next to current, get auto summary from original chat added there.
Start new chat from issue (github or linear) would be nice
These updates to voice input are present when using the Agents Window, but not with chats being used in the main Cursor window. It would be good for those improvements to voice input to be present everywhere voice input is used, so that it is consistent across the application.
That said, I noticed that in the Agents Window, the Ctrl+M shortcut needs to be held, while the shortcut elsewhere is used to toggle voice input off and on, without needing to be held. Certain users will like the shortcut that needs to be held, while others will want the toggle option, and still others may want to flexibly use one method or the other depending on their current situation. So all that needs to be configurable. Don’t keep that flexibility in mind and you will be pushing a portion of your user base to use outside tools for voice input.
This upgrade seems to have broken the PR view in Agents window when you are in Workspace for that repo.
I clearly have a pull request open for this branch but the viewer isn’t picking it up unless I open the repo without any specific workspace. When clicking “Create PR” the agent just responds that one already exists but the Pull Request view in Agents Window remains empty.
Love the 3.1 drop — tiled Agents is super handy when you’re comparing a few runs without tab-hopping.
The Agents window voice upgrade feels like the right direction: hold to talk, full clip, then batch STT with clear UI (waveform / timer / cancel) — basically “finish your thought, then transcribe,” which is what you want for long prompts.
What I’m still hoping for is that same polish in the editor chat, where voice can still feel like a different flow — less natural for back-and-forth (add another take, refine the same message) compared to an agent thread.
A few things that would make it feel one product everywhere:
Same upgraded voice pipeline in editor chat (capture + STT + recording UX as in Agents).
Easier follow-ups — append another voice/text chunk without the “single fragile take” feeling.
Draft/queue the next prompt while something runs so you’re not stuck only hitting stop to plan what’s next.
Cheers — excited to see voice keep converging across surfaces.
Hi, as I mentioned to the Sam support agent AI thingy: I love the agents window, since I am only working with the codebese through the chat.
But I would like to see the option to use “Auto” (for obvious cost saving reasons) on the web version, the cursor.com/agents. Because the only reason why I even use the app, is that mode in particular.