Cursor introduced voice chat with agents on November 6, 2026. A maker can speak to an agent, hear a reply, hand off coding work and ask for progress while that work continues. Cursor says the call works in the Agents Window, the IDE, on the web and in its iOS app. This is a new conversation interface for an existing coding workflow, not a claim that speaking a request makes the agent's code correct. VIBPS has not tested the feature firsthand.
What happens during a voice call with a coding agent?
Cursor describes a two-part interaction. A voice agent handles the spoken conversation and passes implementation work to the coding agent in the chat. While the coding agent runs, the maker can ask what changed in a diff, check progress or discuss an earlier part of the session. The coding agent can speak short updates and a final summary. The useful question for a maker is whether the handoff preserves the exact requirement: after the call, inspect the written task, the diff and the app behavior before treating a spoken summary as completion.
Cursor's release says a call can start from an idle agent with an empty chat input using the Start voice chat control. The documentation requires a signed-in Cursor account and microphone permission; an API-key session cannot start a call. On iOS, Cursor also describes a Siri phrase to begin a new call and Lock Screen controls to mute or end it. These are access paths to a conversation, not a substitute for code review, device testing or checking which project the agent is editing.
How is voice chat different from dictation?
Dictation turns speech into text in the chat input; the maker reviews or sends that text as a normal message. Voice chat is a live, two-way conversation that can continue while the coding agent works. Cursor says its older Voice Mode name referred to dictation, which can make the two easy to confuse. The difference matters when a maker needs a precise instruction: a dictated prompt is visible for editing before sending, while a call invites back-and-forth clarification and progress questions.
What should makers check before relying on it?
Cursor says voice chat uses Grok Voice and that Privacy Mode calls have zero data retention. That is the company's stated behavior; makers working on confidential apps should review the current privacy controls and their organization's rules before discussing private data aloud. More generally, a voice call makes it easier to request a change away from the keyboard, but the review still happens in the project: inspect the files, run the app and confirm the result. VIBPS's separate maker guide covers how to save a recoverable project state before giving any coding agent broad file access.
VIBPS is the social network for software, where the finished app and the maker can meet an audience. Voice can make the making process more accessible, including for someone who wants to talk through an idea while moving around. It does not establish that a new app is ready for people to use. The maker still owns the description, screenshots, access and follow-up after sharing it.