正文 · AI 翻译
Thanks to IT Home reader 惊讶仙人 for the tip!
IT Home reported on November 7 that AI coding tool Cursor announced yesterday (November 6) the launch of its “Voice Chat with Agents” feature.
Developers can now have two-way voice conversations with Cursor’s coding agents, such as verbally describing code changes, asking about task progress, checking diffs and run status, while the agent continues coding in the background.
The feature is available in Cursor’s Agents Window, IDE, web interface, and iOS app. When the agent is idle and the input box is empty, users can click “Start voice chat” to begin a call. On iPhone, they can also start a call by saying “Hey Siri, call Cursor,” and mute or end the call from the lock screen or Dynamic Island.
Unlike the earlier “dictation” feature, which only converted speech to text, live voice chat is a two-way, speech-to-speech conversation. The voice agent listens, responds, and receives tasks, then hands the specific coding work to the coding agent in the current chat; the latter proactively gives brief progress updates while working and a short summary of the result when finished.
During a call, users can also ask about code diffs, check run status, search the agent’s past messages and call history, or switch between agents in a Project with multiple workers. These queries are read-only and do not interrupt coding tasks in progress.
Technically, voice chat is powered by Grok Voice, which handles speech recognition and synthesis; the coding agent continues to use the model selected by the user. Cursor offers several voice options, including Ara, Eve, Leo, Rex, and Sal, and supports switching voices during a call.
On privacy, with Privacy Mode enabled, voice chat will have zero data retention. Cursor also notes that calls have a daily usage limit; if a user remains silent for a long time and the agent has no task running, the call will end automatically after about 10 minutes.