Two-way voice chat + permanent chat delete (not only archive)

Ideas Feature Requests
,
1

Feature request for product/service

Cursor IDE

Describe the request

I’d like two improvements to Cursor chat:

  1. Two-way voice chat
    Right now I can dictate prompts with the mic / Voice Mode, but the agent only replies in text. I’d like a true conversation mode like a phone call or ChatGPT Voice:
  • Speak naturally to the agent
  • Hear the agent’s reply out loud (TTS)
  • Continue turn by turn without typing or reading every response

This would help with hands-free use while walking, commuting, or planning away from the keyboard.

  1. Permanently delete chats
    Right now I can only archive chats. Archive helps declutter, but the chats still remain. Please add a real delete option for:
  • Accidental / test chats
  • Sensitive prompts I don’t want kept
  • Old experiments that are no longer useful

Suggested behavior:

  • “Delete chat” with confirmation
  • Also allow delete from Archive
  • Ideally support bulk delete

Operating System (if it applies)

Windows 10/11

5

Hey @Nuttapong_Yuyang, thanks for the suggestions!

On deleting chats: the Agents Window only offers Archive right now. That’s intentional, so history isn’t lost by accident. If you switch to the Editor window, you can delete individual chats from the chat history there.

We already have a thread collecting votes for Delete in the Agents Window, so I’d suggest adding your voice there! Restore Delete option alongside Archive in chat history

On two-way voice chat: this is something the team is actively looking at, so keep an eye on upcoming releases! For now, voice input is dictation, and replies come back as text.

7

Thanks Colin! I’ll use the Editor window to delete chats for now and upvote that thread. Looking forward to two-way voice chat.

9

+1 — please prioritize true two-way voice conversation for the desktop Agent.

Today mic / Ctrl+M only does dictation; Agent replies stay text-only. I need a phone-call style loop:

1. Speak naturally (Chinese + English)

2. Hear Agent replies via TTS

3. Continue turn-by-turn hands-free

4. Keep full repo/codebase awareness during the voice session

Use case: long Windows coding sessions where typing every instruction is inconvenient. Dictation alone is not enough.

OS: Windows 10/11