Illustrative image: a headset with microphone. Cursor published no official visual for voice chat, so this stands in for the feature. Image: Richard Wheeler (Zephyris), Wikimedia Commons, CC BY-SA 3.0.
📌 Key Highlights
- Cursor shipped Voice Chat with Agents on 6 November 2026 — a realtime, two-way voice call with the coding agent in your chat.
- Talk while the agent works: describe a change out loud, ask what changed in a diff, or check on a run — no typing needed.
- It works in the Agents Window, the IDE, on the web and in Cursor for iOS; on iPhone you can start with "Hey Siri, call Cursor".
- Voice runs on Grok Voice, with voices including Ara, Eve, Leo, Rex and Sal. The coding agent keeps the model you picked for it.
- Limits to know: a signed-in Cursor account is required, a daily call limit applies, and calls end after about 10 minutes of silence. With Privacy Mode on, voice chat runs with zero data retention.
What happened?
Cursor's changelog for 6 November 2026 leads with Voice Chat with Agents: you can now talk with your agents in a realtime voice call. Describe a change out loud, ask what an agent is working on, and hear updates as the work moves forward. It is the second notable Cursor release in two days — it arrived a day after Build Live Apps in Cursor (5 November), which lets a cloud agent build, host and share a web app from a description.
The important distinction is that voice chat is an interaction surface, not a model upgrade. Nothing about the model doing your coding changes: Grok Voice handles listening and speaking, while the coding agent in your chat keeps using the model you selected for it. Cursor's own docs describe it as speech-to-speech — you speak to a voice agent, it replies in a natural voice, and it hands coding work to the coding agent without you typing a word.
Cursor voice chat with agents: how it works
Starting a call takes four steps, per Cursor's documentation. Open an agent chat — in the Agents Window any chat works, including a new one; in the IDE the chat needs at least one message. Wait until the agent is idle and clear the chat input, and a waveform button labelled Start voice chat appears where Send usually sits. Click it, allow microphone access when your computer, browser or phone asks, and the call connects — start talking, or the agent will check in after a few seconds.
On iOS there is a hands-free route: say "Hey Siri, call Cursor" with your iPhone unlocked to start a call with a new agent. When iOS cannot show the call on the system call screen, it appears as a Live Activity on the Lock Screen and in the Dynamic Island, where you can mute or hang up.
Voice chat is separate from dictation, and confusing the two is the easiest mistake. Dictation — which earlier releases called Voice Mode — turns your speech into text in the chat input, and you still review and send it yourself. Voice chat is a two-way conversation where the agent answers out loud. In the IDE, Cmd+Shift+Space still toggles dictation; in the Agents Window you can hold Ctrl+M to talk.
| Feature | Voice chat | Dictation |
|---|---|---|
| Button | Waveform, "Start voice chat" | Microphone, "Start voice input" |
| What you get | A spoken reply from the agent | Text in the chat input |
| Who sends the work | The voice agent hands it off | You review and send |
| Shortcut | None | Cmd+Shift+Space in the IDE; hold Ctrl+M in the Agents Window |
| Where it runs | Agents Window, IDE, web, iOS | Agents Window, IDE, iOS |
.png?width=900)
Illustrative image: a code editor window — the surface where Cursor's voice chat lives. Cursor is built on the same editor foundation as Visual Studio Code. Screenshot: Microsoft, Wikimedia Commons, free license.
What you can do during a call
The voice agent is more than a microphone. While the coding agent works, you can keep the conversation going: hand off a task or a follow-up, ask what changed in a diff, check on a run, search earlier messages and calls with that agent, or switch between workers in a Project that has them. Those lookups are read-only — the voice agent reads diffs, runs and agents to tell you what changed and what is still running.
The coding agent also speaks for itself. It shares short progress notes while it works, a brief answer when it finishes, and an update when workers it started report back. If it does not write a spoken update for a turn, the call reads its written reply instead, so you usually hear a short update rather than the full written reply. The call transcript appears in the chat, and cloud agent transcripts sync to your other devices while the call is live — unless Privacy Mode is on.
There are places the button simply does not appear: subagents and Project workers, archived chats, Claude Code chats, and chats owned by another teammate. In the IDE it is also hidden for cloud agents that are still streaming, and when dictation is turned off. If none of those apply, Cursor's docs say to update the app.
Why it matters for developers
Voice is the third way agents are becoming easier to direct. The same week Cursor added talking, the agent world kept moving on other fronts: Anthropic explained how Claude's agent memory improves between sessions, and IntelliJ IDEA 2026.3 shipped with a Cursor extension of its own. The direction is consistent — agents that remember, agents you can talk to, agents embedded in more editors — and developer tooling like Playwright 1.65 is adding agent-friendly workflows at the same pace.
The practical win is for moments when typing is the bottleneck: reviewing a diff while your hands are busy, talking through a fix on a walk with the iOS app, or dictating a change while the agent runs it. There is also a genuine accessibility angle — developers who find long typing sessions difficult can now direct an agent by voice. But keep expectations honest: no source has published measured time savings or productivity gains from voice chat, and the feature does not make the agent smarter — it only changes how you talk to it.
Should you try it?
✅ Yes, if you already use Cursor's agents. It costs nothing extra to try — Cursor's changelog announces no separate price for the feature — and it is one update away. Try it on a small, well-understood fix first: describe the change out loud, then read the transcript in the chat to check the agent heard what you meant. If you work with sensitive code, turn Privacy Mode on first so the call runs with zero data retention.
⏳ Wait if you hit the limits. API-key-authenticated sessions cannot start voice chat at all, and there is a daily call limit whose number Cursor does not publish — if you reach it, Cursor tells you when you can call again. Calls also hang up by default after about 10 minutes of silence while the agent has no work in progress. And if you mostly work in subagents, archived chats or Claude Code chats, the button will not be there at all.
FAQs
How do I start a voice chat in Cursor?
Open an agent chat, wait until the agent is idle, and clear the chat input. A waveform button labelled "Start voice chat" appears where Send usually is — click it and allow microphone access. In the IDE the chat needs at least one message first.
Is Cursor voice chat the same as dictation?
No. Dictation turns your speech into text in the chat input, which you review and send yourself. Voice chat is a two-way conversation: the agent answers out loud, hands work to the coding agent, and keeps talking while the work runs.
Does Cursor voice chat work on iOS?
Yes. Voice chat works in Cursor for iOS, and you can start a call hands-free by saying "Hey Siri, call Cursor" with your iPhone unlocked. Calls appear as a Live Activity on the Lock Screen and in the Dynamic Island, where you can mute or hang up.
Do I need a Cursor account to use voice chat?
Yes. Voice calls require a signed-in Cursor account — sessions authenticated with an API key cannot start them, even though the Start voice chat button is visible. You also need to allow microphone permission when your system or browser asks.
Takeaway
Cursor's voice chat turns the agent from something you type at into something you talk with — while it keeps coding. The limits are real but narrow: sign in, mind the daily cap, and keep Privacy Mode on for sensitive work. Update Cursor and try one spoken request on your next small fix, then check the transcript to see exactly what the agent heard and did.
Comments
Post a Comment