Playbooks · Coding tasks
Cursor agents take voice calls now
Written November 8, 2026
Cursor's Nov 6 release adds two-way voice calls with coding agents on Grok Voice. Where it works, what it costs, and the daily limit.
What shipped
Voice Chat with Agents, in Cursor's November 6 changelog, is a realtime two-way voice call with the agent in your chat. Describe a change out loud, ask what the agent is doing, and hear updates as the work moves forward. It works in the Agents Window, the IDE, on web, and in Cursor for iOS. When the agent is idle and the chat input is empty, a waveform button labelled Start voice chat appears where Send usually is. Click it, allow microphone access, and start talking, or the agent will check in after a few seconds.
On iOS you can also say "Hey Siri, call Cursor" to start a call with a new agent, and mute or hang up from the Lock Screen with a Live Activity. 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.
How it works
The call is speech-to-speech, and the work is split in two. A voice agent handles the spoken conversation, and when you ask for code changes it hands the work to the coding agent already in that chat. The coding agent keeps the model you picked for it. Grok Voice handles listening and speaking only. That split matters: voice chat is an interaction surface, not a model upgrade. The code is still written by the coding agent you chose, at the quality that model gives you.
During a call, the voice agent can hand off work, answer questions about your work, search the agent's history, switch Project workers, and hang up when you say you are done. Its lookups of diffs, runs, and agents are read-only: it can tell you what changed and what is still running, but it cannot change anything by looking. The coding agent speaks for itself too. It shares short progress notes while it works and a brief answer when it is done, so you usually hear a short update instead of the full written reply. If the coding agent does not write a spoken update for a turn, the call reads its written reply instead.
Voice chat is separate from dictation, which earlier releases called Voice Mode. Dictation still turns your speech into text in the chat input, and Cmd+Shift+Space still toggles it in the IDE. Voice chat is a two-way conversation where the agent answers out loud. Pick a voice in Voice settings, from many named voices including Ara, Eve, Leo, Rex, and Sal. With Privacy Mode on, voice chat runs with zero data retention.
The limits
You need a signed-in Cursor account. Sessions authenticated with an API key cannot start voice chat, and the Start voice chat button shows even when you are signed out or using an API key, so a visible button is not proof a call will connect. Sign in with your account, then start the call again. You also need microphone permission, which your operating system or browser may ask for first.
There is a daily limit on voice calls, and Cursor publishes no number for it. If you reach it, Cursor tells you when you can call again. A call hangs up by default after about 10 minutes of silence from you while the agent has no work in progress. The button is hidden in 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. And there are no plan or pricing details in the changelog or the docs page: the release says nothing about which plans include voice chat or what it costs.
What to try first
Start one call on a task you already understand, in the Agents Window, and ask the agent to describe a change out loud before it makes it. Listen for what the voice agent hands off cleanly and what it garbles. Check the transcript in the chat afterwards: the spoken summary is short by design, and the full reply stays in the chat. If you work in noisy places or with long silences, expect the 10-minute hang-up and the daily limit to shape how you use it. Voice is good for steering and checking in. It is a poor way to review a large diff, which is still a job for your eyes.
Limits
- Nothing here reports a voice call. The surfaces, the Grok Voice split, the limits, and the gaps above come from Cursor's changelog and voice-chat docs, and no call was made while writing this page.
- The daily limit exists but its number is unpublished. This page states no number, and no time saving is claimed, since no measurement exists in the sources read.
- No plan or pricing details are in the changelog or the docs page. The page states that gap rather than filling it from a secondary source.
Related playbooks
- Cursor builds live apps from cloud agents
- OpenAI's Agents API: the harness is now hosted
- Claude agents now dream: memory that consolidates
- Zalando's broker hands agents OAuth safely
Sources
Back to: Coding playbooks hub · Also: Cursor builds live apps from cloud agents · Run Sonnet with an Opus advisor in Claude Code · T3 Code: drive your agents from your phone